pip install -r requirements.txt

# Commande pour installer les dépendances
!pip install -r requirements.txt

Requirement already satisfied: pandas in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 1)) (2.2.3)
Requirement already satisfied: numpy==2.1.0 in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 2)) (2.1.0)
Requirement already satisfied: scikit-learn in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 3)) (1.6.1)
Requirement already satisfied: xlrd in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 4)) (2.0.1)
Requirement already satisfied: openpyxl in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 5)) (3.1.5)
Requirement already satisfied: matplotlib in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 6)) (3.10.0)
Requirement already satisfied: seaborn in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 7)) (0.13.2)
Requirement already satisfied: shap in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 8)) (0.46.0)
Requirement already satisfied: xgboost in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 9)) (2.1.4)
Requirement already satisfied: imbalanced-learn in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 10)) (0.13.0)
Requirement already satisfied: nbconvert in ./mvenv/lib/python3.10/site-packages (from -r requirements.txt (line 11)) (7.16.6)
Requirement already satisfied: python-dateutil>=2.8.2 in ./mvenv/lib/python3.10/site-packages (from pandas->-r requirements.txt (line 1)) (2.9.0.post0)
Requirement already satisfied: pytz>=2020.1 in ./mvenv/lib/python3.10/site-packages (from pandas->-r requirements.txt (line 1)) (2025.1)
Requirement already satisfied: tzdata>=2022.7 in ./mvenv/lib/python3.10/site-packages (from pandas->-r requirements.txt (line 1)) (2025.1)
Requirement already satisfied: scipy>=1.6.0 in ./mvenv/lib/python3.10/site-packages (from scikit-learn->-r requirements.txt (line 3)) (1.15.1)
Requirement already satisfied: threadpoolctl>=3.1.0 in ./mvenv/lib/python3.10/site-packages (from scikit-learn->-r requirements.txt (line 3)) (3.5.0)
Requirement already satisfied: joblib>=1.2.0 in ./mvenv/lib/python3.10/site-packages (from scikit-learn->-r requirements.txt (line 3)) (1.4.2)
Requirement already satisfied: et-xmlfile in ./mvenv/lib/python3.10/site-packages (from openpyxl->-r requirements.txt (line 5)) (2.0.0)
Requirement already satisfied: fonttools>=4.22.0 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (4.56.0)
Requirement already satisfied: pyparsing>=2.3.1 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (3.2.1)
Requirement already satisfied: cycler>=0.10 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (0.12.1)
Requirement already satisfied: contourpy>=1.0.1 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (1.3.1)
Requirement already satisfied: packaging>=20.0 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (24.2)
Requirement already satisfied: kiwisolver>=1.3.1 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (1.4.8)
Requirement already satisfied: pillow>=8 in ./mvenv/lib/python3.10/site-packages (from matplotlib->-r requirements.txt (line 6)) (11.1.0)
Requirement already satisfied: slicer==0.0.8 in ./mvenv/lib/python3.10/site-packages (from shap->-r requirements.txt (line 8)) (0.0.8)
Requirement already satisfied: numba in ./mvenv/lib/python3.10/site-packages (from shap->-r requirements.txt (line 8)) (0.61.0)
Requirement already satisfied: tqdm>=4.27.0 in ./mvenv/lib/python3.10/site-packages (from shap->-r requirements.txt (line 8)) (4.67.1)
Requirement already satisfied: cloudpickle in ./mvenv/lib/python3.10/site-packages (from shap->-r requirements.txt (line 8)) (3.1.1)
Requirement already satisfied: nvidia-nccl-cu12 in ./mvenv/lib/python3.10/site-packages (from xgboost->-r requirements.txt (line 9)) (2.25.1)
Requirement already satisfied: sklearn-compat<1,>=0.1 in ./mvenv/lib/python3.10/site-packages (from imbalanced-learn->-r requirements.txt (line 10)) (0.1.3)
Requirement already satisfied: jupyterlab-pygments in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (0.3.0)
Requirement already satisfied: defusedxml in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (0.7.1)
Requirement already satisfied: markupsafe>=2.0 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (3.0.2)
Requirement already satisfied: jinja2>=3.0 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (3.1.5)
Requirement already satisfied: pygments>=2.4.1 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (2.19.1)
Requirement already satisfied: traitlets>=5.1 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (5.14.3)
Requirement already satisfied: nbclient>=0.5.0 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (0.10.2)
Requirement already satisfied: nbformat>=5.7 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (5.10.4)
Requirement already satisfied: jupyter-core>=4.7 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (5.7.2)
Requirement already satisfied: bleach[css]!=5.0.0 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (6.2.0)
Requirement already satisfied: pandocfilters>=1.4.1 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (1.5.1)
Requirement already satisfied: beautifulsoup4 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (4.13.3)
Requirement already satisfied: mistune<4,>=2.0.3 in ./mvenv/lib/python3.10/site-packages (from nbconvert->-r requirements.txt (line 11)) (3.1.2)
Requirement already satisfied: webencodings in ./mvenv/lib/python3.10/site-packages (from bleach[css]!=5.0.0->nbconvert->-r requirements.txt (line 11)) (0.5.1)
Requirement already satisfied: tinycss2<1.5,>=1.1.0 in ./mvenv/lib/python3.10/site-packages (from bleach[css]!=5.0.0->nbconvert->-r requirements.txt (line 11)) (1.4.0)
Requirement already satisfied: platformdirs>=2.5 in ./mvenv/lib/python3.10/site-packages (from jupyter-core>=4.7->nbconvert->-r requirements.txt (line 11)) (4.3.6)
Requirement already satisfied: typing-extensions in ./mvenv/lib/python3.10/site-packages (from mistune<4,>=2.0.3->nbconvert->-r requirements.txt (line 11)) (4.12.2)
Requirement already satisfied: jupyter-client>=6.1.12 in ./mvenv/lib/python3.10/site-packages (from nbclient>=0.5.0->nbconvert->-r requirements.txt (line 11)) (8.6.3)
Requirement already satisfied: fastjsonschema>=2.15 in ./mvenv/lib/python3.10/site-packages (from nbformat>=5.7->nbconvert->-r requirements.txt (line 11)) (2.21.1)
Requirement already satisfied: jsonschema>=2.6 in ./mvenv/lib/python3.10/site-packages (from nbformat>=5.7->nbconvert->-r requirements.txt (line 11)) (4.23.0)
Requirement already satisfied: six>=1.5 in ./mvenv/lib/python3.10/site-packages (from python-dateutil>=2.8.2->pandas->-r requirements.txt (line 1)) (1.17.0)
Requirement already satisfied: soupsieve>1.2 in ./mvenv/lib/python3.10/site-packages (from beautifulsoup4->nbconvert->-r requirements.txt (line 11)) (2.6)
Requirement already satisfied: llvmlite<0.45,>=0.44.0dev0 in ./mvenv/lib/python3.10/site-packages (from numba->shap->-r requirements.txt (line 8)) (0.44.0)
Requirement already satisfied: rpds-py>=0.7.1 in ./mvenv/lib/python3.10/site-packages (from jsonschema>=2.6->nbformat>=5.7->nbconvert->-r requirements.txt (line 11)) (0.23.1)
Requirement already satisfied: attrs>=22.2.0 in ./mvenv/lib/python3.10/site-packages (from jsonschema>=2.6->nbformat>=5.7->nbconvert->-r requirements.txt (line 11)) (25.1.0)
Requirement already satisfied: referencing>=0.28.4 in ./mvenv/lib/python3.10/site-packages (from jsonschema>=2.6->nbformat>=5.7->nbconvert->-r requirements.txt (line 11)) (0.36.2)
Requirement already satisfied: jsonschema-specifications>=2023.03.6 in ./mvenv/lib/python3.10/site-packages (from jsonschema>=2.6->nbformat>=5.7->nbconvert->-r requirements.txt (line 11)) (2024.10.1)
Requirement already satisfied: pyzmq>=23.0 in ./mvenv/lib/python3.10/site-packages (from jupyter-client>=6.1.12->nbclient>=0.5.0->nbconvert->-r requirements.txt (line 11)) (26.2.1)
Requirement already satisfied: tornado>=6.2 in ./mvenv/lib/python3.10/site-packages (from jupyter-client>=6.1.12->nbclient>=0.5.0->nbconvert->-r requirements.txt (line 11)) (6.4.2)

import warnings
warnings.filterwarnings("ignore")

# Importation des bibliothèques principales
  
import numpy as np  # Manipulation de tableaux numériques, gestion des calculs mathématiques  
import pandas as pd  # Gestion et manipulation des données sous forme de DataFrame  

# Visualisation des données  
import seaborn as sns  # Création de graphiques statistiques avancés  
import matplotlib.pyplot as plt  # Génération de visualisations basiques (histogrammes, scatter plots, etc.)  

# Préparation des données et séparation en train/test  
from sklearn.model_selection import train_test_split  # Découpage des données en ensembles d'entraînement et de test  
from sklearn.preprocessing import OneHotEncoder  # Encodage des variables catégorielles en variables numériques  

# Modèle de base (Régression Logistique)  
from sklearn.linear_model import LogisticRegression  # Modèle de régression logistique pour la classification  

# Évaluation des modèles  
from sklearn.metrics import classification_report  # Génération d’un rapport de performance des modèles  

# Gestion des classes déséquilibrées  
from sklearn.utils import resample  # Sous-échantillonnage de la classe majoritaire (undersampling)  
from imblearn.over_sampling import SMOTE  # Sur-échantillonnage de la classe minoritaire (oversampling)  

# Modèles avancés  
from sklearn.ensemble import RandomForestClassifier  # Modèle de Random Forest pour la classification  
from xgboost import XGBClassifier  # Modèle XGBoost pour une classification optimisée  

# Optimisation des modèles  
from sklearn.model_selection import GridSearchCV  # Recherche des meilleurs hyperparamètres avec validation croisée  
from imblearn.pipeline import Pipeline as ImbPipeline  # Création d'un pipeline intégrant le prétraitement et le modèle  

# Interprétabilité des modèles  
import shap  # Explication des prédictions des modèles avec SHAP (SHapley Additive exPlanations)  

# Profilage
from tqdm import tqdm  # Affichage de barres de progression pour suivre l’exécution des 

import matplotlib.gridspec as gridspec

import plotly.express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
from sklearn.preprocessing import StandardScaler
import xgboost as xgb
from sklearn.pipeline import Pipeline

# Résultats des élections présidentielles de 2020 au niveau des comtés
elections_2020 = pd.read_csv('./data/2020_US_County_Level_Presidential_Results.csv')

# Résultats des élections présidentielles de 2008 à 2016 au niveau des comtés
# Ce fichier ne sera utilisé que pour l’analyse exploratoire (comparaison des tendances)
elections_08_16 = pd.read_csv('./data/US_County_Level_Presidential_Results_08-16.csv')

# Données démographiques : Estimations de la population par comté
population = pd.read_excel('./data/PopulationEstimates.xls', engine='xlrd', header=2)

# Données sur l’éducation : niveaux d’études atteints par comté
education = pd.read_excel('./data/Education.xls', engine='xlrd', header=4)

# Données sur la pauvreté : taux de pauvreté par comté
poverty = pd.read_excel('./data/PovertyEstimates.xls', engine='xlrd', header=4)

# Données sur le chômage : taux de chômage par comté
unemployment = pd.read_excel('./data/Unemployment.xls', engine='xlrd', header=4)

# Harmonisation des colonnes pour assurer une cohérence entre les datasets

# Sélection et renommage des colonnes pour les résultats des élections 2020
df_2020 = elections_2020[['county_fips', 'county_name', 'state_name']].rename(
    columns={'county_fips': 'fips', 'county_name': 'county_name', 'state_name': 'state_name'}
)

# Sélection et renommage des colonnes pour les résultats des élections 2008-2016
df_08_16 = elections_08_16[['fips_code', 'county']].rename(
    columns={'fips_code': 'fips', 'county': 'county_name'}
)

# Sélection et renommage des colonnes pour les données de population
df_population = population[['FIPStxt', 'Area_Name', 'State']].rename(
    columns={'FIPStxt': 'fips', 'Area_Name': 'county_name', 'State': 'state_code'}
)

# Sélection et renommage des colonnes pour les données d'éducation
df_education = education[['FIPS Code', 'Area name', 'State']].rename(
    columns={'FIPS Code': 'fips', 'Area name': 'county_name', 'State': 'state_code'}
)

# Sélection et renommage des colonnes pour les données de pauvreté
df_poverty = poverty[['FIPStxt', 'Area_name', 'Stabr']].rename(
    columns={'FIPStxt': 'fips', 'Area_name': 'county_name', 'Stabr': 'state_code'}
)

# Sélection et renommage des colonnes pour les données de chômage
df_unemployment = unemployment[['fips_txt', 'area_name', 'Stabr']].rename(
    columns={'fips_txt': 'fips', 'area_name': 'county_name', 'Stabr': 'state_code'}
)


# Concaténation de tous les jeux de données en un seul DataFrame
checkpoint_0_raw = pd.concat([
    df_2020,
    df_08_16,
    df_population,
    df_education,
    df_poverty,
    df_unemployment
], ignore_index=True)

checkpoint_0_raw

# Harmonisation du code FIPS sur 5 caractères (ajout de zéros devant si nécessaire)
checkpoint_0_raw['fips'] = checkpoint_0_raw['fips'].astype(str).str.zfill(5)
checkpoint_0_raw

# Suppression des doublons basés sur le FIPS (identifiant unique des comtés)
checkpoint_0 = checkpoint_0_raw.drop_duplicates(subset=['fips']).sort_values(by='fips').reset_index(drop=True)

# Vérification des premières lignes du DataFrame final
checkpoint_0

# Sauvegarde du DataFrame harmonisé dans un fichier Excel
checkpoint_0.to_excel('checkpoints/save_0.xlsx', index=False)
print("\n✅ Données harmonisées sauvegardées dans 'checkpoints/save_0.xlsx'.")

✅ Données harmonisées sauvegardées dans 'checkpoints/save_0.xlsx'.

# Chargement du fichier harmonisé précédent
ch0 = pd.read_excel('checkpoints/save_0.xlsx')  

# Liste des sources utilisées pour compléter les valeurs manquantes
sources = [df_2020, df_08_16, df_population, df_education, df_poverty, df_unemployment]

# Liste des colonnes à compléter si elles sont absentes
columns_to_fill = ['county_name', 'state_code', 'state_name']

# Vérification et création des colonnes manquantes si elles ne sont pas présentes dans `dm0`
for col in columns_to_fill:
    if col not in ch0.columns:
        ch0[col] = pd.NA  # Remplissage initial avec des valeurs manquantes

# Fonction pour compléter les valeurs manquantes en se basant sur les autres sources
def fill_missing_values(base_df, sources, columns_to_fill):
    """
    Remplit les valeurs manquantes d'un DataFrame en utilisant d'autres sources de données.
    
    - base_df : DataFrame principal contenant des valeurs manquantes
    - sources : Liste des DataFrames sources
    - columns_to_fill : Liste des colonnes à compléter
    """
    for source in sources:
        for col in columns_to_fill:
            if col in source.columns:  # Vérifie si la colonne existe dans la source
                base_df[col] = base_df[col].fillna(
                    base_df['fips'].map(source.set_index('fips')[col])  # Remplissage basé sur la correspondance FIPS
                )
    return base_df

# Application de la fonction pour compléter les valeurs manquantes
checkpoint_1_raw = fill_missing_values(ch0, sources, columns_to_fill)

# Vérification des valeurs manquantes après remplissage
ch1_missing_data = checkpoint_1_raw[columns_to_fill].isnull().sum()
print(f"Valeurs manquantes après remplissage :\n{ch1_missing_data}")

Valeurs manquantes après remplissage :
county_name      0
state_code      41
state_name     172
dtype: int64

# Sauvegarde du fichier complété
checkpoint_1_raw.to_excel('checkpoints/save_1.xlsx', index=False)
print("\n✅ Données complétées sauvegardées dans 'checkpoints/save_1.xlsx'.")

✅ Données complétées sauvegardées dans 'checkpoints/save_1.xlsx'.

df_2020 = elections_2020[['county_fips', 'county_name', 'state_name']].rename(
    columns={'county_fips': 'county_code', 'county_name': 'county_name', 'state_name': 'state_name'})

df_08_16 = elections_08_16[['fips_code', 'county']].rename(
    columns={'fips_code': 'county_code', 'county': 'county_name'})

df_population = population[['FIPStxt', 'Area_Name', 'State']].rename(
    columns={'FIPStxt': 'county_code', 'Area_Name': 'county_name', 'State': 'state_code'})

df_education = education[['FIPS Code', 'Area name', 'State']].rename(
    columns={'FIPS Code': 'county_code', 'Area name': 'county_name', 'State': 'state_code'})

df_poverty = poverty[['FIPStxt', 'Area_name', 'Stabr']].rename(
    columns={'FIPStxt': 'county_code', 'Area_name': 'county_name', 'Stabr': 'state_code'})

df_unemployment = unemployment[['fips_txt', 'area_name', 'Stabr']].rename(
    columns={'fips_txt': 'county_code', 'area_name': 'county_name', 'Stabr': 'state_code'})

# Liste des sources et leur priorité
sources = [df_2020, df_08_16, df_population, df_education, df_poverty, df_unemployment]

# Charger le fichier harmonisé existant ou créer une base vide
ch1 = pd.read_excel('checkpoints/save_1.xlsx')

# Renommer la colonne principale en `county_code`
ch1.rename(columns={'fips': 'county_code'}, inplace=True)

# S'assurer que `county_code` est formaté sur 5 chiffres
ch1['county_code'] = ch1['county_code'].astype(str).str.zfill(5)

# Ajouter les colonnes manquantes
columns_to_fill = ['county_name', 'state_code', 'state_name']
for col in columns_to_fill:
    if col not in ch1.columns:
        ch1[col] = pd.NA

# Fonction pour combler les valeurs manquantes
def fill_missing_values(base_df, sources, columns_to_fill):
    for source in sources:
        # S'assurer que county_code est au bon format dans les sources
        source['county_code'] = source['county_code'].astype(str).str.zfill(5)
        for col in columns_to_fill:
            if col in source.columns:  # Vérifier que la colonne existe dans la source
                base_df[col] = base_df[col].fillna(
                    base_df['county_code'].map(source.set_index('county_code')[col])  # Remplir selon la clé county_code
                )
    return base_df

# Compléter les colonnes manquantes
checkpoint_2_raw = fill_missing_values(ch1, sources, columns_to_fill)

# Vérification des valeurs manquantes
ch2_missing_data = checkpoint_2_raw[columns_to_fill].isnull().sum()
print(f"Valeurs manquantes après remplissage :\n{ch2_missing_data}")

Valeurs manquantes après remplissage :
county_name      0
state_code      41
state_name     172
dtype: int64

# Sauvegarder les données complètes
checkpoint_2_raw.to_excel('checkpoints/save_2.xlsx', index=False)
print("\nDonnées complétées sauvegardées dans 'checkpoints/save_2.xlsx'.")

Données complétées sauvegardées dans 'checkpoints/save_2.xlsx'.

elections_2020 = elections_2020[['county_fips', 'per_gop', 'per_dem']].rename(columns={'county_fips': 'county_code'})
elections_08_16 = elections_08_16[['fips_code', 'total_2016', 'dem_2016', 'gop_2016']].rename(columns={'fips_code': 'county_code'})
population = population[['FIPStxt', 'Rural-urban_Continuum Code_2013', 'Urban_Influence_Code_2013']].rename(
    columns={'FIPStxt': 'county_code', 'Rural-urban_Continuum Code_2013': 'rural_urban_code',
             'Urban_Influence_Code_2013': 'urban_influence_code'})
education = education[['FIPS Code', 'Percent of adults with less than a high school diploma, 2015-19',
                       'Percent of adults with a high school diploma only, 2015-19',
                       'Percent of adults completing some college or associate\'s degree, 2015-19',
                       'Percent of adults with a bachelor\'s degree or higher, 2015-19']].rename(
    columns={'FIPS Code': 'county_code',
             'Percent of adults with less than a high school diploma, 2015-19': 'percent_no_highschool',
             'Percent of adults with a high school diploma only, 2015-19': 'percent_highschool',
             'Percent of adults completing some college or associate\'s degree, 2015-19': 'percent_college',
             'Percent of adults with a bachelor\'s degree or higher, 2015-19': 'percent_bachelor'})
poverty = poverty[['FIPStxt', 'PCTPOVALL_2019', 'MEDHHINC_2019']].rename(
    columns={'FIPStxt': 'county_code', 'PCTPOVALL_2019': 'percent_poverty',
             'MEDHHINC_2019': 'median_household_income'})
unemployment = unemployment[['fips_txt', 'Unemployment_rate_2019', 'Employed_2019', 'Unemployed_2019']].rename(
    columns={'fips_txt': 'county_code', 'Unemployment_rate_2019': 'unemployment_rate'})

# Standardiser le format des `county_code` (5 caractères)
datasets = [elections_2020, elections_08_16, population, education, poverty, unemployment]
for df in datasets:
    df['county_code'] = df['county_code'].astype(str).str.zfill(5)

datamap = pd.read_excel('checkpoints/save_2.xlsx')
datamap['county_code'] = datamap['county_code'].astype(str).str.zfill(5)


# Liste des colonnes pertinentes à ajouter
columns_to_add = {
    'elections_2020': ['per_gop', 'per_dem'],
    'elections_08_16': ['total_2016', 'dem_2016', 'gop_2016'],
    'population': ['rural_urban_code', 'urban_influence_code'],
    'education': ['percent_no_highschool', 'percent_highschool', 'percent_college', 'percent_bachelor'],
    'poverty': ['percent_poverty', 'median_household_income'],
    'unemployment': ['unemployment_rate', 'Employed_2019', 'Unemployed_2019']
}

# Compléter les colonnes manquantes dans le fichier datamap
for source, cols in zip(datasets, columns_to_add.values()):
    for col in cols:
        if col not in datamap.columns:
            datamap[col] = pd.NA
        datamap[col] = datamap[col].fillna(datamap['county_code'].map(source.set_index('county_code')[col]))
print(datamap)
# Sauvegarder le fichier enrichi
datamap.to_excel('checkpoints/save_3.xlsx', index=False)
print("\nFichier 'checkpoints/save_3.xlsx' sauvegardé avec toutes les données intégrées.")

     county_code                       county_name state_name state_code  \
0          00000                     United States        NaN         US   
1          01000                           Alabama        NaN         AL   
2          01001                    Autauga County    Alabama         AL   
3          01003                    Baldwin County    Alabama         AL   
4          01005                    Barbour County    Alabama         AL   
...          ...                               ...        ...        ...   
3319       72145  Vega Baja Municipio, Puerto Rico        NaN         PR   
3320       72147    Vieques Municipio, Puerto Rico        NaN         PR   
3321       72149   Villalba Municipio, Puerto Rico        NaN         PR   
3322       72151    Yabucoa Municipio, Puerto Rico        NaN         PR   
3323       72153      Yauco Municipio, Puerto Rico        NaN         PR   

       per_gop   per_dem  total_2016  dem_2016  gop_2016  rural_urban_code  \
0          NaN       NaN         NaN       NaN       NaN               NaN   
1          NaN       NaN         NaN       NaN       NaN               NaN   
2     0.714368  0.270184     24661.0    5908.0   18110.0               2.0   
3     0.761714  0.224090     94090.0   18409.0   72780.0               3.0   
4     0.534512  0.457882     10390.0    4848.0    5431.0               6.0   
...        ...       ...         ...       ...       ...               ...   
3319       NaN       NaN         NaN       NaN       NaN               1.0   
3320       NaN       NaN         NaN       NaN       NaN               7.0   
3321       NaN       NaN         NaN       NaN       NaN               2.0   
3322       NaN       NaN         NaN       NaN       NaN               1.0   
3323       NaN       NaN         NaN       NaN       NaN               2.0   

      urban_influence_code  percent_no_highschool  percent_highschool  \
0                      NaN              11.998918           26.956844   
1                      NaN              13.819302           30.800268   
2                      2.0              11.483395           33.588459   
3                      2.0               9.193843           27.659616   
4                      6.0              26.786907           35.604542   
...                    ...                    ...                 ...   
3319                   1.0              28.428238           26.225822   
3320                  12.0              28.773281           39.177906   
3321                   2.0              21.993263           38.366028   
3322                   1.0              29.048897           25.715004   
3323                   2.0              26.556698           33.272095   

      percent_college  percent_bachelor  percent_poverty  \
0           28.898697         32.145542             12.3   
1           29.912098         25.468332             15.6   
2           28.356571         26.571573             12.1   
3           31.284081         31.862459             10.1   
4           26.029837         11.578713             27.1   
...               ...               ...              ...   
3319        24.123638         21.222300              NaN   
3320        14.049454         17.999357              NaN   
3321        19.727892         19.912819              NaN   
3322        27.233078         18.003019              NaN   
3323        15.529844         24.641363              NaN   

      median_household_income  unemployment_rate  Employed_2019  \
0                     65712.0           3.669409    157115247.0   
1                     51771.0           3.000000      2174483.0   
2                     58233.0           2.700000        25458.0   
3                     59871.0           2.700000        94675.0   
4                     35972.0           3.800000         8213.0   
...                       ...                ...            ...   
3319                      NaN           9.600000        11791.0   
3320                      NaN           6.900000         2406.0   
3321                      NaN          15.900000         6231.0   
3322                      NaN          13.100000         7552.0   
3323                      NaN          14.600000         8331.0   

      Unemployed_2019  
0           5984808.0  
1             67264.0  
2               714.0  
3              2653.0  
4               324.0  
...               ...  
3319           1246.0  
3320            179.0  
3321           1175.0  
3322           1139.0  
3323           1428.0  

[3324 rows x 20 columns]

Fichier 'checkpoints/save_3.xlsx' sauvegardé avec toutes les données intégrées.

# Vérifier les valeurs manquantes
missing_data = datamap.isnull().sum()
print("Valeurs manquantes par colonne :\n", missing_data)

# Pourcentage de valeurs manquantes
missing_percentage = (missing_data / len(datamap)) * 100
print("\nPourcentage de valeurs manquantes par colonne :\n", missing_percentage)

# Vérifier les doublons
duplicate_count = datamap.duplicated().sum()
print(f"Nombre de doublons dans le DataFrame : {duplicate_count}")

# Statistiques descriptives
stats = datamap.describe()
print("Statistiques descriptives :\n", stats)

Valeurs manquantes par colonne :
 county_code                  0
county_name                  0
state_name                 172
state_code                  41
per_gop                    172
per_dem                    172
total_2016                 212
dem_2016                   212
gop_2016                   212
rural_urban_code           104
urban_influence_code       104
percent_no_highschool       51
percent_highschool          51
percent_college             51
percent_bachelor            51
percent_poverty            131
median_household_income    131
unemployment_rate           52
Employed_2019               52
Unemployed_2019             52
dtype: int64

Pourcentage de valeurs manquantes par colonne :
 county_code                0.000000
county_name                0.000000
state_name                 5.174489
state_code                 1.233454
per_gop                    5.174489
per_dem                    5.174489
total_2016                 6.377858
dem_2016                   6.377858
gop_2016                   6.377858
rural_urban_code           3.128761
urban_influence_code       3.128761
percent_no_highschool      1.534296
percent_highschool         1.534296
percent_college            1.534296
percent_bachelor           1.534296
percent_poverty            3.941035
median_household_income    3.941035
unemployment_rate          1.564380
Employed_2019              1.564380
Unemployed_2019            1.564380
dtype: float64
Nombre de doublons dans le DataFrame : 0

Statistiques descriptives :
            per_gop      per_dem    total_2016      dem_2016       gop_2016  \
count  3152.000000  3152.000000  3.112000e+03  3.112000e+03    3112.000000   
mean      0.647805     0.333851  4.089631e+04  1.956104e+04   19343.762211   
std       0.162014     0.159852  1.082522e+05  6.847899e+04   39125.598644   
min       0.053973     0.030909  6.400000e+01  4.000000e+00      57.000000   
25%       0.554128     0.209978  4.815000e+03  1.164750e+03    3206.000000   
50%       0.681720     0.300235  1.092950e+04  3.140000e+03    7113.000000   
75%       0.773776     0.425830  2.866450e+04  9.535250e+03   17391.750000   
max       0.961818     0.921497  2.314275e+06  1.654626e+06  590465.000000   

       rural_urban_code  urban_influence_code  percent_no_highschool  \
count       3220.000000           3220.000000            3273.000000   
mean           4.937888              5.188820              13.330532   
std            2.724344              3.506848               6.545762   
min            1.000000              1.000000               1.116910   
25%            2.000000              2.000000               8.540109   
50%            6.000000              5.000000              11.884497   
75%            7.000000              8.000000              17.020765   
max            9.000000             12.000000              73.560211   

       percent_highschool  percent_college  percent_bachelor  percent_poverty  \
count         3273.000000      3273.000000       3273.000000      3193.000000   
mean            33.956041        30.587866         22.125561        14.417946   
std              7.212828         5.340745          9.536379         5.769337   
min              7.265136         5.235602          0.000000         2.700000   
25%             29.369493        27.022669         15.511985        10.400000   
50%             34.249691        30.628832         19.776859        13.400000   
75%             38.947308        34.079266         26.348045        17.400000   
max             57.433674        60.563381         77.557411        47.700000   

       median_household_income  unemployment_rate  Employed_2019  \
count              3193.000000        3272.000000   3.272000e+03   
mean              55874.761979           4.139630   1.446578e+05   
std               14493.345229           1.785734   2.808116e+06   
min               24732.000000           0.700000   2.120000e+02   
25%               46309.000000           3.000000   4.778750e+03   
50%               53505.000000           3.700000   1.129300e+04   
75%               62327.000000           4.700000   3.218725e+04   
max              151806.000000          19.300000   1.571152e+08   

       Unemployed_2019  
count     3.272000e+03  
mean      5.541794e+03  
std       1.071326e+05  
min       4.000000e+00  
25%       2.030000e+02  
50%       4.990000e+02  
75%       1.328500e+03  
max       5.984808e+06

# Étape initiale : Identifier les états
datamap['is_state'] = datamap['county_code'].apply(lambda x: 1 if x.endswith('000') else 0)
datamap_complete = datamap.copy()

# Supprimer la ligne correspondant à county_code = "00000" (United States)
datamap_complete = datamap_complete[datamap_complete['county_code'] != '00000']
print(f"Ligne avec county_code '00000' (United States) supprimée. Nombre de lignes restantes : {len(datamap_complete)}")

# Mettre à jour state_name pour les états
datamap_complete.loc[datamap_complete['is_state'] == 1, 'state_name'] = datamap_complete['county_name']

# Afficher le dataframe initial
print(datamap_complete)

# Sauvegarder le fichier enrichi
datamap_complete.to_excel('checkpoints/save_4.xlsx', index=False)
print("\nFichier 'checkpoints/save_4.xlsx' sauvegardé avec toutes les données intégrées.")

# Ajouter state_prefix
datamap_complete['state_prefix'] = datamap_complete['county_code'].str[:2]

# Agrégation existante pour per_gop, per_dem, etc.
state_agg = datamap_complete[datamap_complete['is_state'] == 0].groupby('state_prefix').agg({
    'per_gop': 'mean', 
    'per_dem': 'mean', 
    'gop_2016': 'sum', 
    'dem_2016': 'sum', 
    'total_2016': 'sum'
}).reset_index()

# Mettre à jour les lignes des états avec les agrégations
for index, row in state_agg.iterrows():
    state_prefix = row['state_prefix']
    state_county_code = state_prefix + '000'
    datamap_complete.loc[(datamap_complete['is_state'] == 1) & (datamap_complete['county_code'] == state_county_code), 'per_gop'] = row['per_gop']
    datamap_complete.loc[(datamap_complete['is_state'] == 1) & (datamap_complete['county_code'] == state_county_code), 'per_dem'] = row['per_dem']
    datamap_complete.loc[(datamap_complete['is_state'] == 1) & (datamap_complete['county_code'] == state_county_code), 'total_2016'] = row['total_2016']
    datamap_complete.loc[(datamap_complete['is_state'] == 1) & (datamap_complete['county_code'] == state_county_code), 'gop_2016'] = row['gop_2016']
    datamap_complete.loc[(datamap_complete['is_state'] == 1) & (datamap_complete['county_code'] == state_county_code), 'dem_2016'] = row['dem_2016']

# Nouvelle étape : Calculer la distribution des rural_urban_code par état
ruc_dist = datamap_complete[datamap_complete['is_state'] == 0].groupby('state_prefix')['rural_urban_code'].value_counts(normalize=True).unstack(fill_value=0)
ruc_dist.columns = [f'ruc_{int(col)}' for col in ruc_dist.columns]  # Renommer colonnes : ruc_1, ruc_2, etc.

# Si urban_influence_code est présent (préparation pour le futur)
if 'urban_influence_code' in datamap_complete.columns:
    uic_dist = datamap_complete[datamap_complete['is_state'] == 0].groupby('state_prefix')['urban_influence_code'].value_counts(normalize=True).unstack(fill_value=0)
    uic_dist.columns = [f'uic_{int(col)}' for col in uic_dist.columns]  # Renommer : uic_1, uic_2, etc.
else:
    print("Note : 'urban_influence_code' n'est pas dans le dataset. Seuls les RUCC seront calculés.")

# Fusionner avec les lignes des états
df_states = datamap_complete[datamap_complete['is_state'] == 1].set_index('state_prefix')
df_states = df_states.join(ruc_dist, how='left')
if 'urban_influence_code' in datamap_complete.columns:
    df_states = df_states.join(uic_dist, how='left')

# Réintégrer dans le dataframe complet
datamap_complete = pd.concat([datamap_complete[datamap_complete['is_state'] == 0], df_states.reset_index()], ignore_index=True)

# Remplir les NaN avec 0 pour les nouvelles colonnes
for col in ruc_dist.columns:
    datamap_complete[col] = datamap_complete[col].fillna(0)
if 'urban_influence_code' in datamap_complete.columns:
    for col in uic_dist.columns:
        datamap_complete[col] = datamap_complete[col].fillna(0)

# Afficher les colonnes pertinentes, y compris les nouvelles distributions RUCC
print(datamap_complete[['county_code', 'county_name', 'state_name', 'state_code', 'per_gop', 'per_dem', 'total_2016', 'gop_2016', 'dem_2016', 'is_state'] + [col for col in datamap_complete.columns if col.startswith('ruc_')]])

# Sauvegarder le fichier enrichi
datamap_complete.to_excel('checkpoints/save_5.xlsx', index=False)
print("\nFichier 'checkpoints/save_5.xlsx' sauvegardé avec toutes les données intégrées, y compris les distributions RUCC.")

Ligne avec county_code '00000' (United States) supprimée. Nombre de lignes restantes : 3323
     county_code                       county_name state_name state_code  \
1          01000                           Alabama    Alabama         AL   
2          01001                    Autauga County    Alabama         AL   
3          01003                    Baldwin County    Alabama         AL   
4          01005                    Barbour County    Alabama         AL   
5          01007                       Bibb County    Alabama         AL   
...          ...                               ...        ...        ...   
3319       72145  Vega Baja Municipio, Puerto Rico        NaN         PR   
3320       72147    Vieques Municipio, Puerto Rico        NaN         PR   
3321       72149   Villalba Municipio, Puerto Rico        NaN         PR   
3322       72151    Yabucoa Municipio, Puerto Rico        NaN         PR   
3323       72153      Yauco Municipio, Puerto Rico        NaN         PR   

       per_gop   per_dem  total_2016  dem_2016  gop_2016  rural_urban_code  \
1          NaN       NaN         NaN       NaN       NaN               NaN   
2     0.714368  0.270184     24661.0    5908.0   18110.0               2.0   
3     0.761714  0.224090     94090.0   18409.0   72780.0               3.0   
4     0.534512  0.457882     10390.0    4848.0    5431.0               6.0   
5     0.784263  0.206983      8748.0    1874.0    6733.0               1.0   
...        ...       ...         ...       ...       ...               ...   
3319       NaN       NaN         NaN       NaN       NaN               1.0   
3320       NaN       NaN         NaN       NaN       NaN               7.0   
3321       NaN       NaN         NaN       NaN       NaN               2.0   
3322       NaN       NaN         NaN       NaN       NaN               1.0   
3323       NaN       NaN         NaN       NaN       NaN               2.0   

      ...  percent_no_highschool  percent_highschool  percent_college  \
1     ...              13.819302           30.800268        29.912098   
2     ...              11.483395           33.588459        28.356571   
3     ...               9.193843           27.659616        31.284081   
4     ...              26.786907           35.604542        26.029837   
5     ...              20.942602           44.878773        23.800098   
...   ...                    ...                 ...              ...   
3319  ...              28.428238           26.225822        24.123638   
3320  ...              28.773281           39.177906        14.049454   
3321  ...              21.993263           38.366028        19.727892   
3322  ...              29.048897           25.715004        27.233078   
3323  ...              26.556698           33.272095        15.529844   

      percent_bachelor  percent_poverty  median_household_income  \
1            25.468332             15.6                  51771.0   
2            26.571573             12.1                  58233.0   
3            31.862459             10.1                  59871.0   
4            11.578713             27.1                  35972.0   
5            10.378526             20.3                  47918.0   
...                ...              ...                      ...   
3319         21.222300              NaN                      NaN   
3320         17.999357              NaN                      NaN   
3321         19.912819              NaN                      NaN   
3322         18.003019              NaN                      NaN   
3323         24.641363              NaN                      NaN   

      unemployment_rate  Employed_2019  Unemployed_2019  is_state  
1                   3.0      2174483.0          67264.0         1  
2                   2.7        25458.0            714.0         0  
3                   2.7        94675.0           2653.0         0  
4                   3.8         8213.0            324.0         0  
5                   3.1         8419.0            266.0         0  
...                 ...            ...              ...       ...  
3319                9.6        11791.0           1246.0         0  
3320                6.9         2406.0            179.0         0  
3321               15.9         6231.0           1175.0         0  
3322               13.1         7552.0           1139.0         0  
3323               14.6         8331.0           1428.0         0  

[3323 rows x 21 columns]

Fichier 'checkpoints/save_4.xlsx' sauvegardé avec toutes les données intégrées.
     county_code     county_name     state_name state_code   per_gop  \
0          01001  Autauga County        Alabama         AL  0.714368   
1          01003  Baldwin County        Alabama         AL  0.761714   
2          01005  Barbour County        Alabama         AL  0.534512   
3          01007     Bibb County        Alabama         AL  0.784263   
4          01009   Blount County        Alabama         AL  0.895716   
...          ...             ...            ...        ...       ...   
3318       53000      Washington     Washington         WA  0.520402   
3319       54000   West Virginia  West Virginia         WV  0.741402   
3320       55000       Wisconsin      Wisconsin         WI  0.564259   
3321       56000         Wyoming        Wyoming         WY  0.750912   
3322       72000     Puerto Rico    Puerto Rico         PR       NaN   

       per_dem  total_2016   gop_2016   dem_2016  is_state     ruc_1  \
0     0.270184     24661.0    18110.0     5908.0         0  0.000000   
1     0.224090     94090.0    72780.0    18409.0         0  0.000000   
2     0.457882     10390.0     5431.0     4848.0         0  0.000000   
3     0.206983      8748.0     6733.0     1874.0         0  0.000000   
4     0.095694     25384.0    22808.0     2150.0         0  0.000000   
...        ...         ...        ...        ...       ...       ...   
3318  0.448434   2765627.0  1043648.0  1523720.0         1  0.128205   
3319  0.243346    708226.0   486198.0   187457.0         1  0.018182   
3320  0.419635   2944620.0  1409467.0  1382210.0         1  0.097222   
3321  0.217684    248742.0   174248.0    55949.0         1  0.000000   
3322       NaN         0.0        0.0        0.0         1  0.512821   

         ruc_2     ruc_3     ruc_4     ruc_5     ruc_6     ruc_7     ruc_8  \
0     0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   
1     0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   
2     0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   
3     0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   
4     0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   
...        ...       ...       ...       ...       ...       ...       ...   
3318  0.179487  0.230769  0.153846  0.051282  0.102564  0.025641  0.076923   
3319  0.090909  0.272727  0.036364  0.018182  0.236364  0.127273  0.127273   
3320  0.111111  0.152778  0.097222  0.000000  0.277778  0.083333  0.111111   
3321  0.000000  0.086957  0.043478  0.086957  0.043478  0.565217  0.000000   
3322  0.205128  0.166667  0.038462  0.000000  0.051282  0.012821  0.000000   

         ruc_9  
0     0.000000  
1     0.000000  
2     0.000000  
3     0.000000  
4     0.000000  
...        ...  
3318  0.051282  
3319  0.072727  
3320  0.069444  
3321  0.173913  
3322  0.012821  

[3323 rows x 19 columns]

Fichier 'checkpoints/save_5.xlsx' sauvegardé avec toutes les données intégrées, y compris les distributions RUCC.

# Filtrer uniquement les États (is_state == 1)
datamap_states = datamap_complete[datamap_complete['is_state'] == 1].copy()

# Vérification
print(f"Nombre total d'observations : {len(datamap)}")
print(f"Nombre d'États : {len(datamap_states)}")
print(datamap_states.head())

# Sauvegarder le fichier enrichi
datamap_states.to_excel('checkpoints/states_1.xlsx', index=False)
print("\nFichier 'checkpoints/states_1.xlsx' sauvegardé avec toutes les données intégrées.")

Nombre total d'observations : 3324
Nombre d'États : 52
     county_code county_name  state_name state_code   per_gop   per_dem  \
3271       01000     Alabama     Alabama         AL  0.647359  0.342648   
3272       02000      Alaska      Alaska         AK  0.497797  0.420912   
3273       04000     Arizona     Arizona         AZ  0.548723  0.435861   
3274       05000    Arkansas    Arkansas         AR  0.688531  0.282032   
3275       06000  California  California         CA  0.439389  0.537068   

      total_2016   dem_2016   gop_2016  rural_urban_code  ...     uic_3  \
3271   2078165.0   718084.0  1306925.0               NaN  ...  0.044776   
3272         0.0        0.0        0.0               NaN  ...  0.000000   
3273   2062810.0   936250.0  1021154.0               NaN  ...  0.066667   
3274   1108615.0   378729.0   677904.0               NaN  ...  0.040000   
3275   9631972.0  5931283.0  3184721.0               NaN  ...  0.017241   

         uic_4     uic_5     uic_6     uic_7     uic_8     uic_9    uic_10  \
3271  0.059701  0.104478  0.208955  0.029851  0.000000  0.000000  0.044776   
3272  0.000000  0.000000  0.000000  0.034483  0.068966  0.000000  0.068966   
3273  0.066667  0.133333  0.066667  0.000000  0.066667  0.066667  0.000000   
3274  0.013333  0.053333  0.213333  0.040000  0.133333  0.146667  0.066667   
3275  0.051724  0.068966  0.086207  0.034483  0.051724  0.000000  0.000000   

        uic_11    uic_12  
3271  0.074627  0.000000  
3272  0.344828  0.379310  
3273  0.000000  0.000000  
3274  0.026667  0.000000  
3275  0.034483  0.017241  

[5 rows x 43 columns]

Fichier 'checkpoints/states_1.xlsx' sauvegardé avec toutes les données intégrées.

to_drop_state_cols = ['rural_urban_code', 'urban_influence_code','is_state','state_prefix','county_name']
cleaned_state_df = datamap_states.drop(to_drop_state_cols, axis=1)
cleaned_state_df.rename(columns={'county_code': 'id'}, inplace=True)
cleaned_state_df.head()

# Retirons l'état à l'id 72000 qui correspond au 'Puerto Rico' vu qu'il ne dispose pas de details permettant de determiner la variable cible

cleaned_state_df = cleaned_state_df[cleaned_state_df['id'] != '72000']
cleaned_state_df

# Sauvegarder le fichier enrichi
cleaned_state_df.to_excel('checkpoints/states_df.xlsx', index=False)
print("\nFichier 'checkpoints/states_df.xlsx' sauvegardé avec toutes les données intégrées.")

Fichier 'checkpoints/states_df.xlsx' sauvegardé avec toutes les données intégrées.

df = pd.read_excel("checkpoints/states_df.xlsx")

df["target"] = (df["per_gop"] > df["per_dem"]).astype(int)
df = df.drop(columns=["is_state"], errors="ignore")

df.head()

df.shape

(51, 39)

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 51 entries, 0 to 50
Data columns (total 39 columns):
 #   Column                   Non-Null Count  Dtype  
---  ------                   --------------  -----  
 0   id                       51 non-null     int64  
 1   state_name               51 non-null     object 
 2   state_code               51 non-null     object 
 3   per_gop                  51 non-null     float64
 4   per_dem                  51 non-null     float64
 5   total_2016               51 non-null     int64  
 6   dem_2016                 51 non-null     int64  
 7   gop_2016                 51 non-null     int64  
 8   percent_no_highschool    51 non-null     float64
 9   percent_highschool       51 non-null     float64
 10  percent_college          51 non-null     float64
 11  percent_bachelor         51 non-null     float64
 12  percent_poverty          51 non-null     float64
 13  median_household_income  51 non-null     int64  
 14  unemployment_rate        51 non-null     float64
 15  Employed_2019            51 non-null     int64  
 16  Unemployed_2019          51 non-null     int64  
 17  ruc_1                    51 non-null     float64
 18  ruc_2                    51 non-null     float64
 19  ruc_3                    51 non-null     float64
 20  ruc_4                    51 non-null     float64
 21  ruc_5                    51 non-null     float64
 22  ruc_6                    51 non-null     float64
 23  ruc_7                    51 non-null     float64
 24  ruc_8                    51 non-null     float64
 25  ruc_9                    51 non-null     float64
 26  uic_1                    51 non-null     float64
 27  uic_2                    51 non-null     float64
 28  uic_3                    51 non-null     float64
 29  uic_4                    51 non-null     float64
 30  uic_5                    51 non-null     float64
 31  uic_6                    51 non-null     float64
 32  uic_7                    51 non-null     float64
 33  uic_8                    51 non-null     float64
 34  uic_9                    51 non-null     float64
 35  uic_10                   51 non-null     float64
 36  uic_11                   51 non-null     float64
 37  uic_12                   51 non-null     float64
 38  target                   51 non-null     int64  
dtypes: float64(29), int64(8), object(2)
memory usage: 15.7+ KB

df.describe()

plt.figure(figsize=(12, 6))
sns.histplot(df['per_gop'], bins=30, color='blue', label='Votes Républicains', kde=True)
sns.histplot(df['per_dem'], bins=30, color='red', label='Votes Democrates', kde=True)
plt.title('Distribution des votes pour Démocrates et Republicains')
plt.xlabel('Nombre de Votes')
plt.ylabel('Fréquence')
plt.legend()
plt.show()

plt.figure(figsize=(10, 6))
sns.boxplot(x='target', y='unemployment_rate', data=df)
plt.title('Boxplot de taux de chômage par parti politique')
plt.xlabel('Target Variable (0 = Démocrates, 1 = Républicains)')
plt.ylabel('Taux de chômage')
plt.show()

colors = {'dem': '#3333FF', 'gop': '#FF3333'}

# Création d'une figure pour regrouper les analyses univariées
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)

# 1. Variables politiques - Histogrammes
ax1 = fig.add_subplot(gs[0, 0])
sns.histplot(df['per_gop'], bins=30, color=colors['gop'], kde=True, ax=ax1)
ax1.set_title('Distribution des votes Républicains par état', fontsize=14)
ax1.set_xlabel('Pourcentage de votes Républicains (%)')
ax1.set_ylabel('Nombre d\'états')

ax2 = fig.add_subplot(gs[0, 1])
sns.histplot(df['per_dem'], bins=30, color=colors['dem'], kde=True, ax=ax2)
ax2.set_title('Distribution des votes Démocrates par état', fontsize=14)
ax2.set_xlabel('Pourcentage de votes Démocrates (%)')
ax2.set_ylabel('Nombre d\'états')

Text(0, 0.5, "Nombre d'états")

# 2. Variables éducatives - Histogrammes
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)

ax3 = fig.add_subplot(gs[1, 0])
education_vars = ['percent_no_highschool', 'percent_highschool', 
                 'percent_college', 'percent_bachelor']
education_labels = ['Sans diplôme', 'Diplôme secondaire', 
                   'Études supérieures', 'Licence ou plus']
colors_edu = ['#FF9999', '#99FF99', '#9999FF', '#FFFF99']

for i, (var, label, color) in enumerate(zip(education_vars, education_labels, colors_edu)):
    sns.histplot(df[var], bins=20, kde=True, color=color, alpha=0.7, 
                label=label, ax=ax3)
ax3.set_title('Distribution des niveaux d\'éducation par état', fontsize=14)
ax3.set_xlabel('Pourcentage de la population (%)')
ax3.set_ylabel('Nombre d\'états')
ax3.legend()

<matplotlib.legend.Legend at 0x7fdac5e76320>

# 3. Variables économiques - Histogrammes
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)
ax4 = fig.add_subplot(gs[1, 1])
sns.histplot(df['unemployment_rate'], bins=20, kde=True, color='#66CCFF', ax=ax4)
ax4.set_title('Distribution du taux de chômage par état', fontsize=14)
ax4.set_xlabel('Taux de chômage (%)')
ax4.set_ylabel('Nombre d\'états')

ax5 = fig.add_subplot(gs[2, 0])
sns.histplot(df['percent_poverty'], bins=20, kde=True, color='#FF6666', ax=ax5)
ax5.set_title('Distribution du taux de pauvreté par état', fontsize=14)
ax5.set_xlabel('Taux de pauvreté (%)')
ax5.set_ylabel('Nombre d\'états')

ax6 = fig.add_subplot(gs[2, 1])
sns.histplot(df['median_household_income'], bins=20, kde=True, color='#66CC66', ax=ax6)
ax6.set_title('Distribution du revenu médian des ménages par état', fontsize=14)
ax6.set_xlabel('Revenu médian ($)')
ax6.set_ylabel('Nombre d\'états')

Text(0, 0.5, "Nombre d'états")

# 4. Variables géographiques - Distribution des codes RUCC (Rural-Urban Continuum Code)
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)
ax7 = fig.add_subplot(gs[3, 0])
ruc_columns = [col for col in df.columns if col.startswith('ruc_')]
ruc_means = df[ruc_columns].mean().sort_values(ascending=False)
ruc_categories = ['Métropolitain, >1M', 'Métropolitain, 250k-1M', 'Métropolitain, <250k',
                 'Urbain, >20k, adj. métro', 'Urbain, >20k, non-adj. métro',
                 'Urbain, 2.5k-20k, adj. métro', 'Urbain, 2.5k-20k, non-adj. métro',
                 'Rural, adj. métro', 'Rural, non-adj. métro']
colors_ruc = plt.cm.Spectral(np.linspace(0, 1, len(ruc_means)))

ax7.pie(ruc_means, labels=[f"RUC {i+1}" for i in range(len(ruc_means))], 
       autopct='%1.1f%%', startangle=90, colors=colors_ruc)
ax7.set_title('Distribution des codes Rural-Urban Continuum (RUC)', fontsize=14)

Text(0.5, 1.0, 'Distribution des codes Rural-Urban Continuum (RUC)')

# 5. Variables géographiques - Distribution des codes UIC (Urban Influence Codes)
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)
ax8 = fig.add_subplot(gs[3, 1])
uic_columns = [col for col in df.columns if col.startswith('uic_')]
uic_means = df[uic_columns].mean().sort_values(ascending=False)
colors_uic = plt.cm.tab20(np.linspace(0, 1, len(uic_means)))

ax8.pie(uic_means, labels=[f"UIC {i+1}" for i in range(len(uic_means))], 
       autopct='%1.1f%%', startangle=90, colors=colors_uic)
ax8.set_title('Distribution des codes Urban Influence (UIC)', fontsize=14)

Text(0.5, 1.0, 'Distribution des codes Urban Influence (UIC)')

fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)
ax9 = fig.add_subplot(gs[4, 0])

# Utilisation correcte de `palette` dans `sns.boxplot`
sns.boxplot(
    x='target', 
    y='unemployment_rate', 
    data=df, 
    hue='target',
    legend=False, 
    ax=ax9
)

ax9.set_title('Taux de chômage par affiliation politique', fontsize=14)
ax9.set_xlabel('Affiliation politique (0 = Démocrates, 1 = Républicains)')
ax9.set_ylabel('Taux de chômage (%)')

plt.show()

# 2. Boxplots des variables d'éducation par parti
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)
education_fig, education_axes = plt.subplots(2, 2, figsize=(15, 12))
education_axes = education_axes.flatten()

for i, (var, label) in enumerate(zip(education_vars, education_labels)):
    sns.boxplot(x='target', y=var, data=df, hue='target', ax=education_axes[i])
    education_axes[i].set_title(f'{label} par affiliation politique', fontsize=12)
    education_axes[i].set_xlabel('Affiliation politique (0 = Démocrates, 1 = Républicains)')
    education_axes[i].set_ylabel(f'Pourcentage de {label.lower()} (%)')

education_fig.tight_layout()
fig.add_subplot(gs[4, 1])
plt.close(education_fig)
fig.add_subplot(gs[4, 1]).imshow(np.array(education_fig.canvas.renderer._renderer))
fig.add_subplot(gs[4, 1]).axis('off')

(np.float64(0.0), np.float64(1.0), np.float64(0.0), np.float64(1.0))

# 3. Boxplots des variables économiques par parti
fig = plt.figure(figsize=(20, 30))
gs = gridspec.GridSpec(6, 2, figure=fig)
ax10 = fig.add_subplot(gs[5, 0])
sns.boxplot(x='target', y='percent_poverty', data=df, hue='target',
           palette={0: colors['dem'], 1: colors['gop']}, ax=ax10)
ax10.set_title('Taux de pauvreté par affiliation politique', fontsize=14)
ax10.set_xlabel('Affiliation politique (0 = Démocrates, 1 = Républicains)')
ax10.set_ylabel('Taux de pauvreté (%)')

ax11 = fig.add_subplot(gs[5, 1])
sns.boxplot(x='target', y='median_household_income',hue='target',palette={0: colors['dem'], 1: colors['gop']}, data=df, ax=ax11)
ax11.set_title('Revenu médian des ménages par affiliation politique', fontsize=14)
ax11.set_xlabel('Affiliation politique (0 = Démocrates, 1 = Républicains)')
ax11.set_ylabel('Revenu médian ($)')

fig.tight_layout(pad=3.0)
plt.show()

df.columns

Index(['id', 'state_name', 'state_code', 'per_gop', 'per_dem', 'total_2016',
       'dem_2016', 'gop_2016', 'percent_no_highschool', 'percent_highschool',
       'percent_college', 'percent_bachelor', 'percent_poverty',
       'median_household_income', 'unemployment_rate', 'Employed_2019',
       'Unemployed_2019', 'ruc_1', 'ruc_2', 'ruc_3', 'ruc_4', 'ruc_5', 'ruc_6',
       'ruc_7', 'ruc_8', 'ruc_9', 'uic_1', 'uic_2', 'uic_3', 'uic_4', 'uic_5',
       'uic_6', 'uic_7', 'uic_8', 'uic_9', 'uic_10', 'uic_11', 'uic_12',
       'target'],
      dtype='object')

# 4. Matrices de corrélation (avec et sans variables socio-économiques)
socio_eco_vars = ['percent_no_highschool', 'percent_highschool', 
                 'percent_college', 'percent_bachelor', 'percent_poverty', 
                 'median_household_income', 'unemployment_rate']

df_corr = df[socio_eco_vars + ['target']]

# Calcul de la corrélation de Pearson
corr_matrix = df_corr.corr()

# Visualisation de la matrice de corrélation sous forme de heatmap
plt.figure(figsize=(10, 8))
sns.heatmap(corr_matrix, annot=True, cmap='coolwarm', vmin=-1, vmax=1)
plt.title('Matrice de Corrélation entre les Variables(sociaux économiques) et la Cible')
plt.show()

plt.figure(figsize=(16, 12))
socio_eco_vars = ['percent_no_highschool', 'percent_highschool', 
                 'percent_college', 'percent_bachelor', 'percent_poverty', 
                 'median_household_income', 'unemployment_rate']

corr_matrix = df[socio_eco_vars].corr()
mask = np.triu(np.ones_like(corr_matrix, dtype=bool))
cmap = sns.diverging_palette(230, 20, as_cmap=True)

sns.heatmap(corr_matrix, mask=mask, cmap=cmap, vmax=1, vmin=-1, center=0,
           square=True, linewidths=.5, annot=True, fmt=".2f")
plt.title('Matrice de corrélation des variables socio-économiques et politiques', fontsize=16)
plt.show()

# 5. Scatter plots des variables les plus corrélées avec l'affiliation politique
plt.figure(figsize=(18, 10))
potential_predictors = ['percent_no_highschool', 'percent_bachelor', 
                       'percent_poverty', 'median_household_income']

for i, var in enumerate(potential_predictors):
    plt.subplot(2, 2, i+1)
    sns.scatterplot(x=var, y='per_gop', data=df, color=colors['gop'], 
                   alpha=0.7, label='Républicains')
    sns.scatterplot(x=var, y='per_dem', data=df, color=colors['dem'], 
                   alpha=0.7, label='Démocrates')
    sns.regplot(x=var, y='per_gop', data=df, color=colors['gop'], 
               scatter=False, ci=None, line_kws={"linestyle": "--"})
    sns.regplot(x=var, y='per_dem', data=df, color=colors['dem'], 
               scatter=False, ci=None, line_kws={"linestyle": "--"})
    plt.title(f'Relation entre {var} et le pourcentage de votes', fontsize=12)
    plt.xlabel(var)
    plt.ylabel('Pourcentage de votes')
    plt.legend()

plt.tight_layout()
plt.show()

# 6. Analyse de la composition rurale/urbaine par parti politique (Regroupement des codes RUC (Rural Urban Codes))
# RUC 1-3: Zone métropolitaines
# RUC 4-7: Zone urbaine
# RUC 8-9: Zone rurale
# Créer des colonnes agrégées pour l'urbanité
df['urban_pct'] = df[['ruc_1', 'ruc_2', 'ruc_3']].sum(axis=1)  # Zones métropolitaines
df['semi_urban_pct'] = df[['ruc_4', 'ruc_5', 'ruc_6', 'ruc_7']].sum(axis=1)  # Zones urbaines
df['rural_pct'] = df[['ruc_8', 'ruc_9']].sum(axis=1)  # Zones rurales

plt.figure(figsize=(14, 7))
rural_vars = ['urban_pct', 'semi_urban_pct', 'rural_pct']
rural_labels = ['Métropolitain', 'Urbain', 'Rural']

for i, (var, label) in enumerate(zip(rural_vars, rural_labels)):
    plt.subplot(1, 3, i+1)
    sns.boxplot(x='target', y=var, data=df, hue='target', palette={0: colors['dem'], 1: colors['gop']})
    plt.title(f'Distribution {label} (RUC) par affiliation politique', fontsize=12)
    plt.xlabel('Affiliation politique (0 = Démocrates, 1 = Républicains)')
    plt.ylabel(f'Pourcentage {label} (%)')

plt.tight_layout()
plt.show()

# 7. Regroupement des codes UIC (Urban Influence Codes)
# UIC 1-2: Grands comtés métropolitains
# UIC 3-7: Comtés métropolitains de petite taille ou sous influence métropolitaine
# UIC 8-12: Comtés non-métropolitains ou ruraux

# Création des variables agrégées pour les UIC
df['large_metro_uic'] = df[['uic_1', 'uic_2']].sum(axis=1)  # Grands comtés métropolitains
df['small_metro_uic'] = df[['uic_3', 'uic_4', 'uic_5', 'uic_6', 'uic_7']].sum(axis=1)  # Petits comtés métropolitains ou sous influence
df['rural_uic'] = df[['uic_8', 'uic_9', 'uic_10', 'uic_11', 'uic_12']].sum(axis=1)  # Comtés ruraux

# Visualisation similaire à celle que vous avez faite pour les RUC
plt.figure(figsize=(14, 7))
uic_vars = ['large_metro_uic', 'small_metro_uic', 'rural_uic']
uic_labels = ['Grands Métropolitains', 'Petits Métropolitains', 'Ruraux']

for i, (var, label) in enumerate(zip(uic_vars, uic_labels)):
    plt.subplot(1, 3, i+1)
    sns.boxplot(x='target', y=var, data=df, hue='target', palette={0: colors['dem'], 1: colors['gop']})
    plt.title(f'{label} (UIC) par affiliation politique', fontsize=12)
    plt.xlabel('Affiliation politique (0 = Démocrates, 1 = Républicains)')
    plt.ylabel(f'Pourcentage {label} (%)')

plt.tight_layout()
plt.show()

# Ajoutez également ces variables à votre analyse de corrélation
# Pour voir leur relation avec la variable cible
all_vars = socio_eco_vars + rural_vars + uic_vars + ['target']
full_corr_matrix = df[all_vars].corr()

# Visualisation de la corrélation des nouvelles variables avec la cible
target_corr = full_corr_matrix['target'].sort_values(ascending=False)
print("Corrélation avec la variable cible:")
print(target_corr)

Corrélation avec la variable cible:
target                     1.000000
percent_college            0.571915
small_metro_uic            0.565129
semi_urban_pct             0.519181
rural_pct                  0.456753
percent_poverty            0.380079
rural_uic                  0.370999
percent_highschool         0.288327
percent_no_highschool      0.129814
unemployment_rate          0.090066
percent_bachelor          -0.599873
median_household_income   -0.659655
large_metro_uic           -0.663431
urban_pct                 -0.663431
Name: target, dtype: float64

# 8. Stacked bar chart de la composition rurale/urbaine par parti (RUC)
df_dem = df[df['target'] == 0]
df_rep = df[df['target'] == 1]

dem_rural_means = [df_dem['urban_pct'].mean(), df_dem['semi_urban_pct'].mean(), df_dem['rural_pct'].mean()]
rep_rural_means = [df_rep['urban_pct'].mean(), df_rep['semi_urban_pct'].mean(), df_rep['rural_pct'].mean()]

plt.figure(figsize=(10, 6))
width = 0.35
x = np.arange(len(rural_labels))

plt.bar(x - width/2, dem_rural_means, width, color=colors['dem'], alpha=0.7, label='Démocrates')
plt.bar(x + width/2, rep_rural_means, width, color=colors['gop'], alpha=0.7, label='Républicains')

plt.xlabel('Type de zone')
plt.ylabel('Pourcentage moyen (%)')
plt.title('Composition metropolitaine moyenne par affiliation politique (RUC)')
plt.xticks(x, rural_labels)
plt.legend()
plt.grid(axis='y', linestyle='--', alpha=0.7)
plt.tight_layout()
plt.show()

# 9. Stacked bar chart de la composition urbaine par parti (UIC)
df_dem = df[df['target'] == 0]
df_rep = df[df['target'] == 1]

dem_urban_means = [df_dem['large_metro_uic'].mean(), df_dem['small_metro_uic'].mean(), df_dem['rural_uic'].mean()]
rep_urban_means = [df_rep['large_metro_uic'].mean(), df_rep['small_metro_uic'].mean(), df_rep['rural_uic'].mean()]

plt.figure(figsize=(10, 6))
width = 0.35
x = np.arange(len(uic_labels))

plt.bar(x - width/2, dem_urban_means, width, color=colors['dem'], alpha=0.7, label='Démocrates')
plt.bar(x + width/2, rep_urban_means, width, color=colors['gop'], alpha=0.7, label='Républicains')

plt.xlabel('Type de zone')
plt.ylabel('Pourcentage moyen (%)')
plt.title('Composition urbaine/rurale moyenne par affiliation politique (UIC)')
plt.xticks(x, uic_labels)
plt.legend()
plt.grid(axis='y', linestyle='--', alpha=0.7)
plt.tight_layout()
plt.show()

# Ajoutez également ces variables à votre analyse de corrélation
# Pour voir leur relation avec la variable cible
all_vars = socio_eco_vars + rural_vars + uic_vars + ['target']
full_corr_matrix = df[all_vars].corr()

# Visualisation de la corrélation des nouvelles variables avec la cible
target_corr = full_corr_matrix['target'].sort_values(ascending=False)
print("Corrélation avec la variable cible:")
print(target_corr)

Corrélation avec la variable cible:
target                     1.000000
percent_college            0.571915
small_metro_uic            0.565129
semi_urban_pct             0.519181
rural_pct                  0.456753
percent_poverty            0.380079
rural_uic                  0.370999
percent_highschool         0.288327
percent_no_highschool      0.129814
unemployment_rate          0.090066
percent_bachelor          -0.599873
median_household_income   -0.659655
large_metro_uic           -0.663431
urban_pct                 -0.663431
Name: target, dtype: float64

# Visualisation de la matrice de corrélation sous forme de heatmap
plt.figure(figsize=(10, 8))
sns.heatmap(full_corr_matrix, annot=True, cmap='coolwarm', vmin=-1, vmax=1)
plt.title('Matrice de Corrélation entre les Variables et la Cible')
plt.show()

df['margin'] = (df['per_gop'] - df['per_dem'])*100
print(df)

# ---- Carte 1: Tendances électorales avec une échelle de couleur continue ----
fig1 = px.choropleth(
    df,
    locations='state_code',
    locationmode='USA-states',
    color='margin',  # Utiliser la marge au lieu d'une valeur binaire
    color_continuous_scale='RdBu_r',  # Rouge pour GOP, Bleu pour DEM
    range_color=[-30, 30],  # Limiter l'échelle pour mieux voir les différences
    scope='usa',
    title='Tendances électorales par état (écart GOP-DEM)',
    labels={'margin': 'Écart GOP-DEM (%)'}
)
fig1.update_layout(
    geo=dict(showframe=False, showcoastlines=True, projection_scale=1.1),
    title_font_size=18,
    margin=dict(t=50, b=0, l=0, r=0)
)

# ---- Carte 2: Pourcentage rural avec une échelle de couleur continue ----
fig2 = px.choropleth(
    df,
    locations='state_code',
    locationmode='USA-states',
    color='rural_pct',
    color_continuous_scale='Greens',
    scope='usa',
    title='Pourcentage de zones rurales par état',
    labels={'rural_pct': '% Rural'}
)
fig2.update_layout(
    geo=dict(showframe=False, showcoastlines=True, projection_scale=1.1),
    title_font_size=18,
    margin=dict(t=50, b=0, l=0, r=0)
)

# ---- Carte 3: Revenu médian des ménages ----
fig3 = px.choropleth(
    df,
    locations='state_code',
    locationmode='USA-states',
    color='median_household_income',
    color_continuous_scale='Viridis',
    scope='usa',
    title='Revenu médian des ménages par état',
    labels={'median_household_income': 'Revenu médian ($)'}
)
fig3.update_layout(
    geo=dict(showframe=False, showcoastlines=True, projection_scale=1.1),
    title_font_size=18,
    margin=dict(t=50, b=0, l=0, r=0)
)

# ---- Carte 4: Taux de chômage ----
fig4 = px.choropleth(
    df,
    locations='state_code',
    locationmode='USA-states',
    color='unemployment_rate',
    color_continuous_scale='Reds',
    scope='usa',
    title='Taux de chômage par état',
    labels={'unemployment_rate': 'Taux de chômage (%)'}
)
fig4.update_layout(
    geo=dict(showframe=False, showcoastlines=True, projection_scale=1.1),
    title_font_size=18,
    margin=dict(t=50, b=0, l=0, r=0)
)

# ---- Carte 5: Niveau d'éducation (pourcentage avec un diplôme universitaire) ----
fig5 = px.choropleth(
    df,
    locations='state_code',
    locationmode='USA-states',
    color='percent_bachelor',
    color_continuous_scale='Blues',
    scope='usa',
    title="Pourcentage de la population avec un diplôme universitaire",
    labels={'percent_bachelor': '% Diplôme universitaire'}
)
fig5.update_layout(
    geo=dict(showframe=False, showcoastlines=True, projection_scale=1.1),
    title_font_size=18,
    margin=dict(t=50, b=0, l=0, r=0)
)

# ---- Carte 6: Taux de pauvreté ----
fig6 = px.choropleth(
    df,
    locations='state_code',
    locationmode='USA-states',
    color='percent_poverty',
    color_continuous_scale='OrRd',
    scope='usa',
    title="Taux de pauvreté par état",
    labels={'percent_poverty': '% Pauvreté'}
)
fig6.update_layout(
    geo=dict(showframe=False, showcoastlines=True, projection_scale=1.1),
    title_font_size=18,
    margin=dict(t=50, b=0, l=0, r=0)
)

# ---- Scatter plot: Relation entre ruralité et vote républicain ----
fig7 = px.scatter(
    df, 
    x='rural_pct',
    y='per_gop',
    size='total_2016',
    color='margin',
    color_continuous_scale='RdBu_r',
    hover_name='state_name',
    title="Relation entre ruralité et vote républicain",
    labels={
        'rural_pct': '% Rural',
        'per_gop': '% Vote républicain',
        'total_2016': 'Total votes 2016',
        'margin': 'Marge GOP-DEM'
    },
    size_max=40
)
fig7.update_layout(
    title_font_size=18,
    xaxis_title_font_size=14,
    yaxis_title_font_size=14,
    coloraxis_colorbar_title_font_size=14
)

# ---- Scatter plot: Relation entre éducation et vote républicain ----
fig8 = px.scatter(
    df,
    x='percent_bachelor',
    y='per_gop',
    size='total_2016',
    color='median_household_income',
    color_continuous_scale='Viridis',
    hover_name='state_name',
    title="Relation entre niveau d'éducation et vote républicain",
    labels={
        'percent_bachelor': '% Diplôme universitaire',
        'per_gop': '% Vote républicain',
        'total_2016': 'Total votes 2016',
        'median_household_income': 'Revenu médian ($)'
    },
    size_max=40
)
fig8.update_layout(
    title_font_size=18,
    xaxis_title_font_size=14,
    yaxis_title_font_size=14,
    coloraxis_colorbar_title_font_size=14
)

# ---- Heatmap: Corrélation entre les variables ----
# Sélection des variables numériques pertinentes
cols_to_corr = ['per_gop', 'per_dem', 'margin', 'rural_pct', 'percent_no_highschool', 
                'percent_bachelor', 'percent_poverty', 'median_household_income', 
                'unemployment_rate']

# Calcul de la matrice de corrélation
corr_matrix = df[cols_to_corr].corr()

# Créer une heatmap de corrélation
fig9 = px.imshow(
    corr_matrix,
    text_auto='.2f',
    color_continuous_scale='RdBu_r',
    title="Matrice de corrélation entre les variables",
    labels=dict(x="Variables", y="Variables", color="Corrélation")
)
fig9.update_layout(
    title_font_size=18,
    xaxis_title_font_size=14,
    yaxis_title_font_size=14
)

# ---- Bar plot: Top 10 des états les plus républicains et démocrates ----
# On Créé un dataframe pour les 10 états les plus républicains et les 10 les plus démocrates
top_gop = df.sort_values('per_gop', ascending=False).head(10)
top_dem = df.sort_values('per_dem', ascending=False).head(10)

fig10 = make_subplots(rows=1, cols=2, subplot_titles=("Top 10 des états républicains", "Top 10 des états démocrates"))

fig10.add_trace(
    go.Bar(
        x=top_gop['state_name'],
        y=top_gop['per_gop'],
        marker_color='red',
        name='% Républicain'
    ),
    row=1, col=1
)

fig10.add_trace(
    go.Bar(
        x=top_dem['state_name'],
        y=top_dem['per_dem'],
        marker_color='blue',
        name='% Démocrate'
    ),
    row=1, col=2
)

fig10.update_layout(
    title_text="Top 10 des états par affiliation politique",
    title_font_size=18,
    showlegend=True,
    height=500
)

# Affichons tous les graphiques
for fig in [fig1, fig2, fig3, fig4, fig5, fig6, fig7, fig8, fig9, fig10]:
    fig.show()

       id            state_name state_code   per_gop   per_dem  total_2016  \
0    1000               Alabama         AL  0.647359  0.342648     2078165   
1    2000                Alaska         AK  0.497797  0.420912           0   
2    4000               Arizona         AZ  0.548723  0.435861     2062810   
3    5000              Arkansas         AR  0.688531  0.282032     1108615   
4    6000            California         CA  0.439389  0.537068     9631972   
5    8000              Colorado         CO  0.559502  0.417248     2564185   
6    9000           Connecticut         CT  0.424576  0.557866     1623542   
7   10000              Delaware         DE  0.443037  0.542737      441535   
8   11000  District of Columbia         DC  0.053973  0.921497      280272   
9   12000               Florida         FL  0.633620  0.357409     9386750   
10  13000               Georgia         GA  0.639809  0.350515     4029564   
11  15000                Hawaii         HI  0.330023  0.648358      428825   
12  16000                 Idaho         ID  0.730509  0.240739      688235   
13  17000              Illinois         IL  0.652193  0.327196     5374280   
14  18000               Indiana         IN  0.688717  0.291546     2722029   
15  19000                  Iowa         IA  0.638318  0.344197     1542880   
16  20000                Kansas         KS  0.752552  0.227409     1147143   
17  21000              Kentucky         KY  0.740423  0.245121     1922218   
18  22000             Louisiana         LA  0.646492  0.339012     2027731   
19  23000                 Maine         ME  0.486435  0.485294      741550   
20  24000              Maryland         MD  0.477075  0.498362     2474543   
21  25000         Massachusetts         MA  0.308799  0.668589     3231531   
22  26000              Michigan         MI  0.596681  0.387700     4789450   
23  27000             Minnesota         MN  0.602136  0.376106     2916404   
24  28000           Mississippi         MS  0.562834  0.423827     1162987   
25  29000              Missouri         MO  0.752084  0.232851     2775098   
26  30000               Montana         MT  0.689422  0.287197      483574   
27  31000              Nebraska         NE  0.780742  0.198169      805638   
28  32000                Nevada         NV  0.696978  0.277140     1122990   
29  33000         New Hampshire         NH  0.458564  0.523999      730628   
30  34000            New Jersey         NJ  0.437915  0.544962     3674893   
31  35000            New Mexico         NM  0.532221  0.447902      783127   
32  36000              New York         NY  0.508439  0.471869     7046175   
33  37000        North Carolina         NC  0.584579  0.403189     4629471   
34  38000          North Dakota         ND  0.725569  0.247834      336968   
35  39000                  Ohio         OH  0.674596  0.310536     5325395   
36  40000              Oklahoma         OK  0.778398  0.202755     1451056   
37  41000                Oregon         OR  0.566156  0.404151     1808575   
38  42000          Pennsylvania         PA  0.635927  0.350978     5970107   
39  44000          Rhode Island         RI  0.380590  0.598526      450121   
40  45000        South Carolina         SC  0.535671  0.452410     2084444   
41  46000          South Dakota         SD  0.673640  0.305517      370047   
42  47000             Tennessee         TN  0.747807  0.237236     2484691   
43  48000                 Texas         TX  0.743895  0.245202     8903237   
44  49000                  Utah         UT  0.728772  0.241405      852461   
45  50000               Vermont         VT  0.351706  0.615269      291413   
46  51000              Virginia         VA  0.551998  0.431626     3844787   
47  53000            Washington         WA  0.520402  0.448434     2765627   
48  54000         West Virginia         WV  0.741402  0.243346      708226   
49  55000             Wisconsin         WI  0.564259  0.419635     2944620   
50  56000               Wyoming         WY  0.750912  0.217684      248742   

    dem_2016  gop_2016  percent_no_highschool  percent_highschool  ...  \
0     718084   1306925              13.819302           30.800268  ...   
1          0         0               7.152934           28.003729  ...   
2     936250   1021154              12.860705           23.858877  ...   
3     378729    677904              13.430243           34.034885  ...   
4    5931283   3184721              16.692171           20.487896  ...   
5    1212209   1137455               8.253678           21.368059  ...   
6     884432    668266               9.369879           26.854712  ...   
7     235581    185103               9.982669           31.292805  ...   
8     260223     11553               9.076816           16.835115  ...   
9    4485745   4605515              11.810859           28.573500  ...   
10   1837300   2068623              12.855142           27.714716  ...   
11    266827    128815               8.028257           27.356155  ...   
12    189677    407199               9.226700           27.356852  ...   
13   2977498   2118179              10.787586           25.954943  ...   
14   1031953   1556220              11.181375           33.406757  ...   
15    650790    798923               7.908792           30.982542  ...   
16    414788    656009               9.048409           25.906137  ...   
17    628834   1202942              13.738696           32.893917  ...   
18    779535   1178004              14.773869           33.962753  ...   
19    354873    334838               7.389770           31.473530  ...   
20   1497951    873646               9.796140           24.611042  ...   
21   1964768   1083069               9.242436           24.019262  ...   
22   2267373   2279210               9.190457           28.873878  ...   
23   1366676   1322891               6.859513           24.647589  ...   
24    462001    678457              15.493731           30.438028  ...   
25   1054889   1585753              10.078580           30.617037  ...   
26    174521    274120               6.449651           28.832081  ...   
27    273858    485819               8.595745           26.106092  ...   
28    537753    511319              13.309174           28.085070  ...   
29    348126    345598               6.894038           27.419645  ...   
30   2021756   1535513              10.183736           27.185795  ...   
31    380724    315875              14.411727           26.460430  ...   
32   4143874   2640570              13.179301           25.977776  ...   
33   2162074   2339603              12.219548           25.652466  ...   
34     93526    216133               7.351314           26.429096  ...   
35   2317001   2771984               9.621357           33.037495  ...   
36    419788    947934              11.976947           31.330032  ...   
37    934631    742506               9.287846           22.735300  ...   
38   2844705   2912941               9.480545           34.693886  ...   
39    249902    179421              11.186517           28.267363  ...   
40    849469   1143611              12.488965           29.103146  ...   
41    117442    227701               8.253245           30.237791  ...   
42    867110   1517402              12.537143           32.088009  ...   
43   3867816   4681590              16.313875           24.957039  ...   
44    237241    397004               7.719078           22.836246  ...   
45    178179     95053               7.327794           28.795351  ...   
46   1916845   1731156              10.305691           23.953545  ...   
47   1523720   1043648               8.672709           21.999466  ...   
48    187457    486198              13.097753           40.320034  ...   
49   1382210   1409467               7.791683           30.638811  ...   
50     55949    174248               6.834035           29.073072  ...   

      uic_11    uic_12  target  urban_pct  semi_urban_pct  rural_pct  \
0   0.074627  0.000000       1   0.432836        0.402985   0.164179   
1   0.344828  0.379310       1   0.103448        0.310345   0.586207   
2   0.000000  0.000000       1   0.533333        0.466667   0.000000   
3   0.026667  0.000000       1   0.266667        0.560000   0.173333   
4   0.034483  0.017241       0   0.637931        0.293103   0.068966   
5   0.125000  0.171875       1   0.265625        0.421875   0.312500   
6   0.000000  0.000000       0   0.875000        0.125000   0.000000   
7   0.000000  0.000000       0   1.000000        0.000000   0.000000   
8   0.000000  0.000000       0   1.000000        0.000000   0.000000   
9   0.000000  0.014925       1   0.656716        0.313433   0.029851   
10  0.012579  0.025157       1   0.465409        0.396226   0.138365   
11  0.000000  0.000000       0   0.600000        0.400000   0.000000   
12  0.045455  0.045455       1   0.272727        0.500000   0.227273   
13  0.019608  0.019608       1   0.392157        0.509804   0.098039   
14  0.000000  0.000000       1   0.478261        0.467391   0.054348   
15  0.050505  0.040404       1   0.212121        0.585859   0.202020   
16  0.066667  0.209524       1   0.180952        0.419048   0.400000   
17  0.041667  0.075000       1   0.291667        0.408333   0.300000   
18  0.031250  0.000000       1   0.546875        0.375000   0.078125   
19  0.187500  0.000000       1   0.312500        0.562500   0.125000   
20  0.000000  0.000000       0   0.791667        0.208333   0.000000   
21  0.071429  0.000000       0   0.785714        0.214286   0.000000   
22  0.108434  0.024096       1   0.313253        0.518072   0.168675   
23  0.034483  0.068966       1   0.310345        0.471264   0.218391   
24  0.000000  0.000000       1   0.207317        0.536585   0.256098   
25  0.034783  0.017391       1   0.295652        0.443478   0.260870   
26  0.160714  0.285714       1   0.089286        0.392857   0.517857   
27  0.086022  0.204301       1   0.139785        0.311828   0.548387   
28  0.117647  0.000000       1   0.235294        0.529412   0.235294   
29  0.000000  0.000000       0   0.300000        0.700000   0.000000   
30  0.000000  0.000000       0   1.000000        0.000000   0.000000   
31  0.090909  0.060606       1   0.212121        0.606061   0.181818   
32  0.000000  0.000000       1   0.612903        0.370968   0.016129   
33  0.010000  0.030000       1   0.460000        0.380000   0.160000   
34  0.037736  0.264151       1   0.113208        0.188679   0.698113   
35  0.000000  0.000000       1   0.431818        0.545455   0.022727   
36  0.012987  0.012987       1   0.233766        0.558442   0.207792   
37  0.055556  0.083333       1   0.361111        0.500000   0.138889   
38  0.000000  0.000000       1   0.552239        0.388060   0.059701   
39  0.000000  0.000000       0   1.000000        0.000000   0.000000   
40  0.000000  0.000000       1   0.565217        0.413043   0.021739   
41  0.030303  0.166667       1   0.121212        0.242424   0.636364   
42  0.000000  0.010526       1   0.442105        0.389474   0.168421   
43  0.051181  0.062992       1   0.322835        0.484252   0.192913   
44  0.103448  0.103448       1   0.344828        0.482759   0.172414   
45  0.071429  0.000000       0   0.214286        0.571429   0.214286   
46  0.007519  0.037594       1   0.601504        0.240602   0.157895   
47  0.000000  0.051282       1   0.538462        0.333333   0.128205   
48  0.000000  0.036364       1   0.381818        0.418182   0.200000   
49  0.027778  0.069444       1   0.361111        0.458333   0.180556   
50  0.130435  0.086957       1   0.086957        0.739130   0.173913   

    large_metro_uic  small_metro_uic  rural_uic     margin  
0          0.432836         0.447761   0.119403  30.471133  
1          0.103448         0.034483   0.862069   7.688521  
2          0.533333         0.333333   0.133333  11.286270  
3          0.266667         0.360000   0.373333  40.649831  
4          0.637931         0.258621   0.103448  -9.767912  
5          0.265625         0.187500   0.546875  14.225424  
6          0.875000         0.125000   0.000000 -13.329063  
7          1.000000         0.000000   0.000000  -9.969955  
8          1.000000         0.000000   0.000000 -86.752373  
9          0.656716         0.328358   0.014925  27.621053  
10         0.465409         0.364780   0.169811  28.929338  
11         0.600000         0.000000   0.400000 -31.833484  
12         0.272727         0.409091   0.318182  48.976967  
13         0.392157         0.362745   0.245098  32.499742  
14         0.478261         0.445652   0.076087  39.717157  
15         0.212121         0.383838   0.404040  29.412134  
16         0.180952         0.200000   0.619048  52.514368  
17         0.291667         0.266667   0.441667  49.530237  
18         0.546875         0.328125   0.125000  30.747987  
19         0.312500         0.500000   0.187500   0.114054  
20         0.791667         0.208333   0.000000  -2.128640  
21         0.785714         0.071429   0.142857 -35.979075  
22         0.313253         0.204819   0.481928  20.898059  
23         0.310345         0.390805   0.298851  22.602929  
24         0.207317         0.365854   0.426829  13.900687  
25         0.295652         0.356522   0.347826  51.923302  
26         0.089286         0.232143   0.678571  40.222544  
27         0.139785         0.193548   0.666667  58.257352  
28         0.235294         0.294118   0.470588  41.983823  
29         0.300000         0.400000   0.300000  -6.543491  
30         1.000000         0.000000   0.000000 -10.704621  
31         0.212121         0.242424   0.545455   8.431905  
32         0.612903         0.322581   0.064516   3.657005  
33         0.460000         0.400000   0.140000  18.138959  
34         0.113208         0.207547   0.679245  47.773558  
35         0.431818         0.500000   0.068182  36.405993  
36         0.233766         0.324675   0.441558  57.564274  
37         0.361111         0.277778   0.361111  16.200463  
38         0.552239         0.343284   0.104478  28.494876  
39         1.000000         0.000000   0.000000 -21.793601  
40         0.565217         0.413043   0.021739   8.326079  
41         0.121212         0.212121   0.666667  36.812326  
42         0.442105         0.410526   0.147368  51.057099  
43         0.322835         0.385827   0.291339  49.869359  
44         0.344828         0.206897   0.448276  48.736691  
45         0.214286         0.357143   0.428571 -26.356277  
46         0.601504         0.285714   0.112782  12.037252  
47         0.538462         0.333333   0.128205   7.196842  
48         0.381818         0.400000   0.218182  49.805581  
49         0.361111         0.486111   0.152778  14.462402  
50         0.086957         0.086957   0.826087  53.322846  

[51 rows x 46 columns]

df["target"].value_counts(normalize=True)

target
1    0.784314
0    0.215686
Name: proportion, dtype: float64

df.columns

Index(['id', 'state_name', 'state_code', 'per_gop', 'per_dem', 'total_2016',
       'dem_2016', 'gop_2016', 'percent_no_highschool', 'percent_highschool',
       'percent_college', 'percent_bachelor', 'percent_poverty',
       'median_household_income', 'unemployment_rate', 'Employed_2019',
       'Unemployed_2019', 'ruc_1', 'ruc_2', 'ruc_3', 'ruc_4', 'ruc_5', 'ruc_6',
       'ruc_7', 'ruc_8', 'ruc_9', 'uic_1', 'uic_2', 'uic_3', 'uic_4', 'uic_5',
       'uic_6', 'uic_7', 'uic_8', 'uic_9', 'uic_10', 'uic_11', 'uic_12',
       'target', 'urban_pct', 'semi_urban_pct', 'rural_pct', 'large_metro_uic',
       'small_metro_uic', 'rural_uic', 'margin'],
      dtype='object')

# Liste des colonnes pertinentes
selected_columns = [
    "id","state_name", "state_code",
    "percent_college", "semi_urban_pct", "median_household_income", 
    "percent_poverty", "rural_pct", "percent_bachelor"
]

selected_features = [col for col in selected_columns if col in df.columns]

# Création des jeux de données
X = df[selected_features]
y = df["target"]

categorical_features = ["id","state_name", "state_code"]
encoder = OneHotEncoder(drop="first", sparse_output=False)
X_encoded = pd.DataFrame(encoder.fit_transform(X[categorical_features]))

X_encoded.columns = encoder.get_feature_names_out(categorical_features)
X = X.drop(columns=categorical_features)

# Concaténation les features encodées avec les autres features
X_final = pd.concat([X, X_encoded], axis=1)
X_final.head()

def model_comparison(X, y):
    # Séparation des données en train et test
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)
    
    # List of models to evaluate
    models = {
        'Logistic Regression': LogisticRegression(max_iter=5000, random_state=42),
        'Random Forest': RandomForestClassifier(random_state=42),
        'XGBoost': xgb.XGBClassifier(random_state=42)
    }

    # Paramètres à tester dans GridSearchCV pour chaque modèle
    param_grids = {
        'Logistic Regression': {
            'logreg__C': [0.1, 1, 10],
            'logreg__solver': ['lbfgs', 'liblinear']
        },
        'Random Forest': {
            'rf__n_estimators': [100, 200, 500],
            'rf__max_depth': [10, 20, None],
            'rf__min_samples_split': [2, 5],
            'rf__min_samples_leaf': [1, 2]
        },
        'XGBoost': {
            'xgb__learning_rate': [0.01, 0.1, 0.3],
            'xgb__max_depth': [3, 6, 10],
            'xgb__n_estimators': [50, 100, 200],
            'xgb__subsample': [0.8, 1.0],
            'xgb__colsample_bytree': [0.8, 1.0]
        }
    }
    
    results = []
    
    for model_name, model in models.items():
        # Création d'un pipeline pour chaque modèle
        if model_name == 'Logistic Regression':
            pipeline = Pipeline([('scaler', StandardScaler()), ('logreg', model)])
        elif model_name == 'Random Forest':
            pipeline = Pipeline([('scaler', StandardScaler()), ('rf', model)])
        else:  # XGBoost
            pipeline = Pipeline([('scaler', StandardScaler()), ('xgb', model)])

        # GridSearchCV
        grid_search = GridSearchCV(pipeline, param_grids[model_name], cv=5, scoring='f1', n_jobs=-1)
        
        # Entraînement et évaluation
        grid_search.fit(X_train, y_train)
        y_pred_train = grid_search.predict(X_train)
        y_pred_test = grid_search.predict(X_test)
        
        # Collecte des résultats
        results.append({
            'Model': model_name,
            'Best Params': grid_search.best_params_,
            'Train F1-Score': classification_report(y_train, y_pred_train, output_dict=True)['1']['f1-score'],
            'Test F1-Score': classification_report(y_test, y_pred_test, output_dict=True)['1']['f1-score'],
            'Train Accuracy': classification_report(y_train, y_pred_train, output_dict=True)['accuracy'],
            'Test Accuracy': classification_report(y_test, y_pred_test, output_dict=True)['accuracy'],
            'Train Recall': classification_report(y_train, y_pred_train, output_dict=True)['1']['recall'],
            'Test Recall': classification_report(y_test, y_pred_test, output_dict=True)['1']['recall'],
            'Train Precision': classification_report(y_train, y_pred_train, output_dict=True)['1']['precision'],
            'Test Precision': classification_report(y_test, y_pred_test, output_dict=True)['1']['precision']
        })
    
    # Conversion des résultats en DataFrame
    results_df = pd.DataFrame(results)
    
    return results_df

def plot_model_metrics(results_df):
    """
    Function to plot test metrics from model comparison results
    """
    models = results_df['Model']
    test_accuracy = results_df['Test Accuracy']
    test_f1 = results_df['Test F1-Score']
    test_recall = results_df['Test Recall']
    
    # Couleurs harmonieuses et moins contrastées
    colors = ['#8ecae6', '#219ebc', '#126782']
    
    x = np.arange(len(models))
    width = 0.25  # Plus étroit pour accommoder 3 barres
    
    plt.figure(figsize=(12, 7))
    
    # Création des barres avec les nouvelles couleurs
    plt.bar(x - width, test_accuracy, width, label='Test Accuracy', color=colors[0], alpha=0.8)
    plt.bar(x, test_f1, width, label='Test F1-Score', color=colors[1], alpha=0.8)
    plt.bar(x + width, test_recall, width, label='Test Recall', color=colors[2], alpha=0.8)
    
    plt.xlabel('Modèles', fontsize=12)
    plt.ylabel('Score', fontsize=12)
    plt.title('Comparaison des métriques de test par modèle', fontsize=14)
    plt.xticks(x, models, rotation=15, ha='right')
    plt.ylim(0, 1.1)
    plt.legend()
    plt.grid(True, linestyle='--', alpha=0.3)
    
    # Ajout des valeurs sur les barres
    for i, v in enumerate(test_accuracy):
        plt.text(i - width, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    for i, v in enumerate(test_f1):
        plt.text(i, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    for i, v in enumerate(test_recall):
        plt.text(i + width, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    
    plt.tight_layout()
    plt.show()

def plot_accuracy_comparison(results_df):
    """
    Function to compare train and test accuracy for each model
    """
    models = results_df['Model']
    train_accuracy = results_df['Train Accuracy']
    test_accuracy = results_df['Test Accuracy']
    
    # Couleurs harmonieuses
    colors = ['#219ebc', '#fb8500']
    
    x = np.arange(len(models))
    width = 0.35
    
    plt.figure(figsize=(12, 7))
    
    plt.bar(x - width/2, train_accuracy, width, label='Train Accuracy', color=colors[0], alpha=0.8)
    plt.bar(x + width/2, test_accuracy, width, label='Test Accuracy', color=colors[1], alpha=0.8)
    
    plt.xlabel('Modèles', fontsize=12)
    plt.ylabel('Accuracy', fontsize=12)
    plt.title('Comparaison de l\'accuracy en train et test par modèle', fontsize=14)
    plt.xticks(x, models, rotation=15, ha='right')
    plt.ylim(0, 1.1)
    plt.legend()
    plt.grid(True, linestyle='--', alpha=0.3)
    
    # Ajout des valeurs sur les barres
    for i, v in enumerate(train_accuracy):
        plt.text(i - width/2, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    for i, v in enumerate(test_accuracy):
        plt.text(i + width/2, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    
    # Calcul et affichage de la différence entre train et test (overfitting)
    for i in range(len(models)):
        diff = train_accuracy[i] - test_accuracy[i]
        plt.text(i, max(train_accuracy[i], test_accuracy[i]) + 0.08, 
                 f'Diff: {diff:.2f}', ha='center', fontsize=10, color='#d62828')
    
    plt.tight_layout()
    plt.show()

def plot_recall_comparison(results_df):
    """
    Function to compare train and test recall for each model
    """
    models = results_df['Model']
    train_recall = results_df['Train Recall']
    test_recall = results_df['Test Recall']
    
    # Couleurs harmonieuses
    colors = ['#4cc9f0', '#f72585']
    
    x = np.arange(len(models))
    width = 0.35
    
    plt.figure(figsize=(12, 7))
    
    plt.bar(x - width/2, train_recall, width, label='Train Recall', color=colors[0], alpha=0.8)
    plt.bar(x + width/2, test_recall, width, label='Test Recall', color=colors[1], alpha=0.8)
    
    plt.xlabel('Modèles', fontsize=12)
    plt.ylabel('Recall', fontsize=12)
    plt.title('Comparaison du recall en train et test par modèle', fontsize=14)
    plt.xticks(x, models, rotation=15, ha='right')
    plt.ylim(0, 1.1)
    plt.legend()
    plt.grid(True, linestyle='--', alpha=0.3)
    
    # Ajout des valeurs sur les barres
    for i, v in enumerate(train_recall):
        plt.text(i - width/2, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    for i, v in enumerate(test_recall):
        plt.text(i + width/2, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    
    # Calcul et affichage de la différence entre train et test
    for i in range(len(models)):
        diff = train_recall[i] - test_recall[i]
        plt.text(i, max(train_recall[i], test_recall[i]) + 0.08, 
                 f'Diff: {diff:.2f}', ha='center', fontsize=10, color='#d62828')
    
    plt.tight_layout()
    plt.show()

def plot_f1_comparison(results_df):
    """
    Function to compare train and test F1-Score for each model
    """
    models = results_df['Model']
    train_f1 = results_df['Train F1-Score']
    test_f1 = results_df['Test F1-Score']
    
    # Couleurs harmonieuses
    colors = ['#2a9d8f', '#e76f51']
    
    x = np.arange(len(models))
    width = 0.35
    
    plt.figure(figsize=(12, 7))
    
    plt.bar(x - width/2, train_f1, width, label='Train F1-Score', color=colors[0], alpha=0.8)
    plt.bar(x + width/2, test_f1, width, label='Test F1-Score', color=colors[1], alpha=0.8)
    
    plt.xlabel('Modèles', fontsize=12)
    plt.ylabel('F1-Score', fontsize=12)
    plt.title('Comparaison du F1-Score en train et test par modèle', fontsize=14)
    plt.xticks(x, models, rotation=15, ha='right')
    plt.ylim(0, 1.1)
    plt.legend()
    plt.grid(True, linestyle='--', alpha=0.3)
    
    # Ajout des valeurs sur les barres
    for i, v in enumerate(train_f1):
        plt.text(i - width/2, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    for i, v in enumerate(test_f1):
        plt.text(i + width/2, v + 0.02, f'{v:.2f}', ha='center', fontsize=9)
    
    # Calcul et affichage de la différence entre train et test
    for i in range(len(models)):
        diff = train_f1[i] - test_f1[i]
        plt.text(i, max(train_f1[i], test_f1[i]) + 0.08, 
                 f'Diff: {diff:.2f}', ha='center', fontsize=10, color='#d62828')
    
    plt.tight_layout()
    plt.show()

# Entrainement des modèles
results = model_comparison(X_final, y)
results

# Affichage des resultats
plot_model_metrics(results)

plot_accuracy_comparison(results)
plot_recall_comparison(results)
plot_f1_comparison(results)

def analyze_feature_importance(X, y, model_results):
    """
    Analyse l'importance des variables de façon globale et locale avec SHAP
    Supporte les modèles binaires et multi-classes
    
    Parameters:
    -----------
    X : DataFrame
        Les features utilisées pour l'entraînement
    y : Series
        La variable cible
    model_results : DataFrame
        Résultats de la fonction model_comparison
        
    Returns:
    --------
    dict: Dictionnaire contenant les résultats de l'analyse
    """
    
    # Récupération du meilleur modèle (Random Forest)
    best_params = model_results.loc[model_results['Model'] == 'Random Forest', 'Best Params'].values[0]
    
    # Extraction des paramètres optimaux pour Random Forest
    params = {}
    for key, value in best_params.items():
        if key.startswith('rf__'):
            params[key[4:]] = value  # Enlever le préfixe 'rf__'
    
    # Création et entraînement du modèle avec les meilleurs paramètres
    best_rf = RandomForestClassifier(**params, random_state=42)
    
    # Création et entraînement du modèle de régression logistique pour comparer
    best_params_logreg = model_results.loc[model_results['Model'] == 'Logistic Regression', 'Best Params'].values[0]
    params_logreg = {}
    for key, value in best_params_logreg.items():
        if key.startswith('logreg__'):
            params_logreg[key[8:]] = value  # Enlever le préfixe 'logreg__'
    
    best_logreg = LogisticRegression(**params_logreg, random_state=42)
    
    # Standardisation des données pour la régression logistique
    scaler = StandardScaler()
    X_scaled = scaler.fit_transform(X)
    
    # Entraînement des modèles
    best_rf.fit(X, y)
    best_logreg.fit(X_scaled, y)
    
    # 1. Analyse globale - Feature Importance pour Random Forest
    plt.figure(figsize=(10, 6))
    feature_importances = pd.DataFrame(
        {'feature': X.columns, 'importance': best_rf.feature_importances_}
    ).sort_values('importance', ascending=False)
    
    sns.barplot(x='importance', y='feature', data=feature_importances, palette='Blues_d')
    plt.title('Importance des variables - Random Forest', fontsize=14)
    plt.xlabel('Importance')
    plt.ylabel('Variables')
    plt.tight_layout()
    plt.show()
    
    # 2. Analyse globale - Coefficients pour Régression Logistique
    plt.figure(figsize=(10, 6))
    coefs = pd.DataFrame(
        {'feature': X.columns, 'coefficient': best_logreg.coef_[0]}
    ).sort_values('coefficient', ascending=False)
    
    # Utiliser une palette de couleurs différentes pour les coefficients positifs et négatifs
    colors = ['#FF9999' if c < 0 else '#66B2FF' for c in coefs['coefficient']]
    
    sns.barplot(x='coefficient', y='feature', data=coefs, palette=colors)
    plt.title('Coefficients - Régression Logistique', fontsize=14)
    plt.axvline(x=0, color='gray', linestyle='--')
    plt.xlabel('Coefficient')
    plt.ylabel('Variables')
    plt.tight_layout()
    plt.show()
    
    # 3. Analyse locale - SHAP pour Random Forest
    # Échantillonnage si le dataset est très grand
    sample_size = min(100, len(X))
    X_sample = X.sample(sample_size, random_state=42)
    
    # Création de l'explainer SHAP
    explainer = shap.TreeExplainer(best_rf)
    
    # Pour les graphiques summary plot, on peut encore utiliser shap_values
    shap_values = explainer.shap_values(X_sample)
    
    # Vérifier si nous avons un modèle binaire
    is_binary = isinstance(shap_values, list) or (isinstance(shap_values, np.ndarray) and shap_values.ndim > 2)
    
    # Résumé des valeurs SHAP - utiliser la classe 1 pour classification binaire
    plt.figure(figsize=(10, 8))
    if is_binary:
        # Pour les modèles binaires, utiliser l'index 1 (classe positive)
        shap.summary_plot(shap_values[1] if isinstance(shap_values, list) else shap_values[:, :, 1], 
                          X_sample, plot_type="bar", show=False)
    else:
        shap.summary_plot(shap_values, X_sample, plot_type="bar", show=False)
    
    plt.title('Résumé des valeurs SHAP - Random Forest', fontsize=14)
    plt.tight_layout()
    plt.show()
    
    # Détail des valeurs SHAP
    plt.figure(figsize=(12, 10))
    if is_binary:
        # Pour les modèles binaires, utiliser l'index 1 (classe positive)
        shap.summary_plot(shap_values[1] if isinstance(shap_values, list) else shap_values[:, :, 1], 
                          X_sample, show=False)
    else:
        shap.summary_plot(shap_values, X_sample, show=False)
    
    plt.title('Impact des variables sur les prédictions (SHAP) - Random Forest', fontsize=14)
    plt.tight_layout()
    plt.show()
    
    # 4. Analyse d'un exemple spécifique
    try:
        # Prendre un exemple au hasard
        example_idx = np.random.randint(0, len(X_sample))
        
        print(f"Analyse d'un exemple spécifique (indice {example_idx}):")
        
        # Afficher les valeurs de l'exemple
        example_data = X_sample.iloc[example_idx]
        print("\nValeurs de l'exemple:")
        for feature, value in example_data.items():
            print(f"  {feature}: {value}")
        
        # Tenter d'utiliser le SHAP Dependence plot au lieu du waterfall
        # qui est plus robuste pour les modèles binaires
        plt.figure(figsize=(10, 6))
        
        # Identifier la feature la plus importante
        if is_binary:
            feature_importance = np.abs(shap_values[1]).mean(0) if isinstance(shap_values, list) else np.abs(shap_values[:, :, 1]).mean(0)
        else:
            feature_importance = np.abs(shap_values).mean(0)
            
        most_important_idx = np.argmax(feature_importance)
        most_important_feature = X.columns[most_important_idx]
        
        # Créer un dependence plot pour la feature la plus importante
        if is_binary:
            shap_values_to_plot = shap_values[1] if isinstance(shap_values, list) else shap_values[:, :, 1]
        else:
            shap_values_to_plot = shap_values
            
        shap.dependence_plot(
            most_important_idx, 
            shap_values_to_plot, 
            X_sample,
            feature_names=X.columns,
            show=False
        )
        
        plt.title(f'Dependence plot pour {most_important_feature}', fontsize=14)
        plt.tight_layout()
        plt.show()
        
        # Créer une fonction pour visualiser les contributions des features pour un exemple
        def plot_feature_contributions(example_idx, shap_values, X_sample, is_binary=False):
            # Extraire les valeurs SHAP pour l'exemple
            if is_binary:
                shap_vals = shap_values[1][example_idx] if isinstance(shap_values, list) else shap_values[example_idx, :, 1]
            else:
                shap_vals = shap_values[example_idx]
                
            # Créer un DataFrame pour la visualisation
            contrib_df = pd.DataFrame({
                'Feature': X_sample.columns,
                'Contribution': shap_vals
            }).sort_values('Contribution', ascending=False)
            
            # Visualiser
            plt.figure(figsize=(10, 6))
            bars = plt.barh(contrib_df['Feature'], contrib_df['Contribution'])
            
            # Colorer les barres en fonction de leur contribution (positive/négative)
            for i, bar in enumerate(bars):
                if contrib_df['Contribution'].iloc[i] > 0:
                    bar.set_color('#66B2FF')  # Bleu pour contributions positives
                else:
                    bar.set_color('#FF9999')  # Rouge pour contributions négatives
            
            plt.axvline(x=0, color='gray', linestyle='--')
            plt.title(f'Contributions des variables pour l\'exemple #{example_idx}', fontsize=14)
            plt.xlabel('Impact sur la prédiction')
            plt.tight_layout()
            plt.show()
            
            # Afficher les contributions numériques
            print("\nContributions des variables (top 5):")
            for _, row in contrib_df.head(5).iterrows():
                print(f"  {row['Feature']}: {row['Contribution']:.4f}")
                
            print("\nContributions des variables (bottom 5):")
            for _, row in contrib_df.tail(5).iterrows():
                print(f"  {row['Feature']}: {row['Contribution']:.4f}")
                
            # Calculer la prédiction
            if is_binary:
                expected_value = explainer.expected_value[1] if isinstance(explainer.expected_value, list) else explainer.expected_value
            else:
                expected_value = explainer.expected_value
                
            prediction = expected_value + np.sum(shap_vals)
            print(f"\nValeur de base (expected value): {expected_value:.4f}")
            print(f"Somme des contributions: {np.sum(shap_vals):.4f}")
            print(f"Prédiction finale: {prediction:.4f}")
            
            if is_binary:
                print(f"Probabilité: {1 / (1 + np.exp(-prediction)):.4f}")
        
        # Utiliser notre fonction personnalisée
        plot_feature_contributions(example_idx, shap_values, X_sample, is_binary)
            
    except Exception as e:
        print(f"Erreur lors de l'analyse de l'exemple: {e}")
    
    print("\nAnalyse de l'importance des variables terminée.")
    
    return {
        "feature_importance_rf": feature_importances,
        "coefficients_logreg": coefs,
        "shap_explainer": explainer,
        "shap_values": shap_values,
        "X_sample": X_sample
    }

importance_analysis = analyze_feature_importance(X, y, results)
importance_analysis

Analyse d'un exemple spécifique (indice 27):

Valeurs de l'exemple:
  percent_college: 29.9120979309082
  semi_urban_pct: 0.4029850746268656
  median_household_income: 51771.0
  percent_poverty: 15.6
  rural_pct: 0.16417910447761197
  percent_bachelor: 25.46833229064941

<Figure size 1000x600 with 0 Axes>

Contributions des variables (top 5):
  rural_pct: 0.0588
  median_household_income: 0.0429
  percent_bachelor: 0.0330
  semi_urban_pct: 0.0325
  percent_college: 0.0297

Contributions des variables (bottom 5):
  median_household_income: 0.0429
  percent_bachelor: 0.0330
  semi_urban_pct: 0.0325
  percent_college: 0.0297
  percent_poverty: 0.0120
Erreur lors de l'analyse de l'exemple: unsupported format string passed to numpy.ndarray.__format__

Analyse de l'importance des variables terminée.

{'feature_importance_rf':                    feature  importance
 4                rural_pct    0.296621
 2  median_household_income    0.192839
 0          percent_college    0.168414
 1           semi_urban_pct    0.164082
 5         percent_bachelor    0.118620
 3          percent_poverty    0.059424,
 'coefficients_logreg':                    feature  coefficient
 0          percent_college     0.338587
 4                rural_pct     0.291271
 1           semi_urban_pct     0.283146
 3          percent_poverty     0.198093
 5         percent_bachelor    -0.249499
 2  median_household_income    -0.361014,
 'shap_explainer': <shap.explainers._tree.TreeExplainer at 0x7fdad4ea3a30>,
 'shap_values': array([[[-5.69003582e-03,  5.69003582e-03],
         [-3.70237548e-02,  3.70237548e-02],
         [-4.29828621e-02,  4.29828621e-02],
         [-1.73212954e-02,  1.73212954e-02],
         [-5.81950436e-02,  5.81950436e-02],
         [-3.76105377e-02,  3.76105377e-02]],
 
        [[-3.23404282e-02,  3.23404282e-02],
         [-3.54733745e-02,  3.54733745e-02],
         [-4.54803791e-02,  4.54803791e-02],
         [-2.04290890e-02,  2.04290890e-02],
         [-2.80710254e-02,  2.80710254e-02],
         [-3.70292332e-02,  3.70292332e-02]],
 
        [[ 1.30980957e-02, -1.30980957e-02],
         [ 1.01373954e-02, -1.01373954e-02],
         [-4.99943063e-02,  4.99943063e-02],
         [-5.31275008e-04,  5.31275008e-04],
         [-9.00251500e-02,  9.00251500e-02],
         [ 1.08491711e-01, -1.08491711e-01]],
 
        [[-5.58919284e-02,  5.58919284e-02],
         [-3.11128012e-02,  3.11128012e-02],
         [-3.87881638e-02,  3.87881638e-02],
         [-2.41225010e-03,  2.41225010e-03],
         [-5.01496435e-02,  5.01496435e-02],
         [-3.04687423e-02,  3.04687423e-02]],
 
        [[-4.34703395e-02,  4.34703395e-02],
         [-2.94851777e-02,  2.94851777e-02],
         [-4.10398173e-02,  4.10398173e-02],
         [-1.00328889e-02,  1.00328889e-02],
         [-4.95695039e-02,  4.95695039e-02],
         [-3.52258022e-02,  3.52258022e-02]],
 
        [[-4.03918529e-02,  4.03918529e-02],
         [-9.73775060e-03,  9.73775060e-03],
         [-4.12205966e-02,  4.12205966e-02],
         [-8.22771787e-03,  8.22771787e-03],
         [-5.34671897e-02,  5.34671897e-02],
         [-3.57784217e-02,  3.57784217e-02]],
 
        [[-3.23246441e-02,  3.23246441e-02],
         [-3.29164854e-02,  3.29164854e-02],
         [-4.26690692e-02,  4.26690692e-02],
         [-1.30768543e-02,  1.30768543e-02],
         [-5.48511427e-02,  5.48511427e-02],
         [-3.29853337e-02,  3.29853337e-02]],
 
        [[ 1.29090915e-01, -1.29090915e-01],
         [-6.44150689e-02,  6.44150689e-02],
         [-4.65608541e-02,  4.65608541e-02],
         [-5.05119758e-02,  5.05119758e-02],
         [-5.07835557e-02,  5.07835557e-02],
         [ 3.43570101e-02, -3.43570101e-02]],
 
        [[-3.24619072e-02,  3.24619072e-02],
         [-3.03973937e-02,  3.03973937e-02],
         [-4.28018291e-02,  4.28018291e-02],
         [-1.15030465e-02,  1.15030465e-02],
         [-5.48845311e-02,  5.48845311e-02],
         [-3.67748218e-02,  3.67748218e-02]],
 
        [[ 1.24248851e-01, -1.24248851e-01],
         [ 1.35224880e-01, -1.35224880e-01],
         [ 1.50157423e-01, -1.50157423e-01],
         [ 3.65254869e-02, -3.65254869e-02],
         [ 2.42168606e-01, -2.42168606e-01],
         [ 1.02851224e-01, -1.02851224e-01]],
 
        [[ 1.36949614e-03, -1.36949614e-03],
         [-5.22828835e-02,  5.22828835e-02],
         [-4.87709754e-02,  4.87709754e-02],
         [-1.37170739e-03,  1.37170739e-03],
         [-7.81701520e-02,  7.81701520e-02],
         [ 2.04026927e-02, -2.04026927e-02]],
 
        [[ 1.30252480e-01, -1.30252480e-01],
         [ 1.47945364e-01, -1.47945364e-01],
         [ 1.57618067e-01, -1.57618067e-01],
         [-2.13457296e-02,  2.13457296e-02],
         [ 2.31761643e-01, -2.31761643e-01],
         [ 1.14944647e-01, -1.14944647e-01]],
 
        [[-4.25333267e-02,  4.25333267e-02],
         [-3.39742395e-02,  3.39742395e-02],
         [-4.00258930e-02,  4.00258930e-02],
         [-4.38139975e-03,  4.38139975e-03],
         [-5.76275179e-02,  5.76275179e-02],
         [-3.02811526e-02,  3.02811526e-02]],
 
        [[ 1.58339954e-01, -1.58339954e-01],
         [ 9.75818029e-02, -9.75818029e-02],
         [ 1.30937814e-01, -1.30937814e-01],
         [ 1.40137791e-02, -1.40137791e-02],
         [ 2.60197755e-01, -2.60197755e-01],
         [ 1.10105366e-01, -1.10105366e-01]],
 
        [[-7.49791350e-02,  7.49791350e-02],
         [-3.83200571e-02,  3.83200571e-02],
         [ 8.96482594e-02, -8.96482594e-02],
         [-6.15822698e-03,  6.15822698e-03],
         [-7.67519150e-02,  7.67519150e-02],
         [ 7.73754529e-03, -7.73754529e-03]],
 
        [[ 4.38430883e-02, -4.38430883e-02],
         [ 9.18289826e-02, -9.18289826e-02],
         [ 2.71347826e-01, -2.71347826e-01],
         [ 1.67548636e-02, -1.67548636e-02],
         [ 3.25221134e-02, -3.25221134e-02],
         [ 7.48795970e-02, -7.48795970e-02]],
 
        [[-3.92724441e-02,  3.92724441e-02],
         [-2.98950497e-02,  2.98950497e-02],
         [-4.11336549e-02,  4.11336549e-02],
         [-1.03783352e-02,  1.03783352e-02],
         [-5.13692238e-02,  5.13692238e-02],
         [-3.67748218e-02,  3.67748218e-02]],
 
        [[-3.15281459e-02,  3.15281459e-02],
         [-3.23469056e-02,  3.23469056e-02],
         [-4.29226776e-02,  4.29226776e-02],
         [-1.35254596e-02,  1.35254596e-02],
         [-6.26290540e-02,  6.26290540e-02],
         [-2.58712866e-02,  2.58712866e-02]],
 
        [[-3.97712179e-02,  3.97712179e-02],
         [-2.59650266e-02,  2.59650266e-02],
         [-4.25257047e-02,  4.25257047e-02],
         [-4.05397639e-03,  4.05397639e-03],
         [-6.51682577e-02,  6.51682577e-02],
         [-2.13393461e-02,  2.13393461e-02]],
 
        [[ 1.19078681e-01, -1.19078681e-01],
         [-4.23660377e-02,  4.23660377e-02],
         [-4.32820707e-02,  4.32820707e-02],
         [-3.74501565e-02,  3.74501565e-02],
         [-5.78672157e-02,  5.78672157e-02],
         [-5.69367293e-02,  5.69367293e-02]],
 
        [[-4.99420666e-02,  4.99420666e-02],
         [-1.01766532e-02,  1.01766532e-02],
         [-3.88150760e-02,  3.88150760e-02],
         [-1.31048728e-04,  1.31048728e-04],
         [-5.56475135e-02,  5.56475135e-02],
         [-3.41111715e-02,  3.41111715e-02]],
 
        [[-3.19006032e-02,  3.19006032e-02],
         [-2.46983026e-02,  2.46983026e-02],
         [-4.79928095e-02,  4.79928095e-02],
         [-2.19612368e-02,  2.19612368e-02],
         [-2.47388986e-02,  2.47388986e-02],
         [-3.75316788e-02,  3.75316788e-02]],
 
        [[-5.31273647e-02,  5.31273647e-02],
         [-3.20356891e-02,  3.20356891e-02],
         [-4.18200530e-02,  4.18200530e-02],
         [-3.87507927e-03,  3.87507927e-03],
         [-7.71999604e-02,  7.71999604e-02],
         [ 9.23461708e-03, -9.23461708e-03]],
 
        [[-4.49079717e-02,  4.49079717e-02],
         [-3.15105770e-02,  3.15105770e-02],
         [-3.93178490e-02,  3.93178490e-02],
         [-1.47056493e-02,  1.47056493e-02],
         [-6.71094962e-02,  6.71094962e-02],
         [-1.12719862e-02,  1.12719862e-02]],
 
        [[-5.80460238e-02,  5.80460238e-02],
         [-4.29680437e-02,  4.29680437e-02],
         [-3.06914329e-02,  3.06914329e-02],
         [ 3.06823174e-02, -3.06823174e-02],
         [-7.85338002e-02,  7.85338002e-02],
         [ 1.07334537e-02, -1.07334537e-02]],
 
        [[-3.49605994e-02,  3.49605994e-02],
         [-3.72765775e-02,  3.72765775e-02],
         [-4.26830748e-02,  4.26830748e-02],
         [-1.20440347e-02,  1.20440347e-02],
         [-5.16592915e-02,  5.16592915e-02],
         [-3.01999514e-02,  3.01999514e-02]],
 
        [[-3.86336028e-02,  3.86336028e-02],
         [-2.75128896e-02,  2.75128896e-02],
         [ 2.15482691e-01, -2.15482691e-01],
         [ 7.66952686e-02, -7.66952686e-02],
         [ 2.98277355e-01, -2.98277355e-01],
         [ 4.68676491e-02, -4.68676491e-02]],
 
        [[-2.96763224e-02,  2.96763224e-02],
         [-3.24664485e-02,  3.24664485e-02],
         [-4.29086720e-02,  4.29086720e-02],
         [-1.19975833e-02,  1.19975833e-02],
         [-5.87891694e-02,  5.87891694e-02],
         [-3.29853337e-02,  3.29853337e-02]],
 
        [[ 2.62401617e-01, -2.62401617e-01],
         [ 1.53281412e-02, -1.53281412e-02],
         [-1.13783404e-02,  1.13783404e-02],
         [ 7.05713637e-02, -7.05713637e-02],
         [-4.41577108e-02,  4.41577108e-02],
         [ 1.48411400e-01, -1.48411400e-01]],
 
        [[-5.70271412e-02,  5.70271412e-02],
         [-2.36582961e-02,  2.36582961e-02],
         [-4.15038688e-02,  4.15038688e-02],
         [-3.48109108e-03,  3.48109108e-03],
         [-6.44714400e-02,  6.44714400e-02],
         [-1.86816922e-02,  1.86816922e-02]],
 
        [[-6.36930468e-02,  6.36930468e-02],
         [ 9.50493382e-02, -9.50493382e-02],
         [-4.77279310e-02,  4.77279310e-02],
         [-5.63373806e-03,  5.63373806e-03],
         [-6.80438124e-02,  6.80438124e-02],
         [-2.87743393e-02,  2.87743393e-02]],
 
        [[-5.34747535e-02,  5.34747535e-02],
         [-3.17162180e-02,  3.17162180e-02],
         [-3.76145233e-02,  3.76145233e-02],
         [ 1.83455960e-02, -1.83455960e-02],
         [-7.89799211e-02,  7.89799211e-02],
         [ 8.46162906e-02, -8.46162906e-02]],
 
        [[ 3.68502375e-02, -3.68502375e-02],
         [-1.10728352e-02,  1.10728352e-02],
         [ 1.20929911e-01, -1.20929911e-01],
         [ 8.35737736e-02, -8.35737736e-02],
         [ 2.89126620e-01, -2.89126620e-01],
         [ 1.11768763e-01, -1.11768763e-01]],
 
        [[-5.80337427e-02,  5.80337427e-02],
         [-3.51713246e-02,  3.51713246e-02],
         [-3.96986603e-02,  3.96986603e-02],
         [-1.96845813e-03,  1.96845813e-03],
         [-7.37627538e-02,  7.37627538e-02],
         [-1.88589976e-04,  1.88589976e-04]],
 
        [[-5.84697061e-02,  5.84697061e-02],
         [-2.33043764e-02,  2.33043764e-02],
         [-7.74383211e-03,  7.74383211e-03],
         [ 3.89604942e-03, -3.89604942e-03],
         [-5.81764540e-02,  5.81764540e-02],
         [-2.50252101e-02,  2.50252101e-02]],
 
        [[ 1.45945307e-01, -1.45945307e-01],
         [ 8.97559251e-02, -8.97559251e-02],
         [ 1.75780297e-01, -1.75780297e-01],
         [ 2.42293696e-02, -2.42293696e-02],
         [ 2.40123137e-01, -2.40123137e-01],
         [ 1.05342435e-01, -1.05342435e-01]],
 
        [[-6.92980370e-02,  6.92980370e-02],
         [-6.00726929e-02,  6.00726929e-02],
         [-4.93514759e-02,  4.93514759e-02],
         [-2.72276985e-02,  2.72276985e-02],
         [ 2.42441525e-01, -2.42441525e-01],
         [-4.53151498e-02,  4.53151498e-02]],
 
        [[ 1.54544597e-01, -1.54544597e-01],
         [ 2.20545756e-01, -2.20545756e-01],
         [ 2.12440261e-03, -2.12440261e-03],
         [ 1.75522539e-02, -1.75522539e-02],
         [ 3.13186693e-01, -3.13186693e-01],
         [ 5.32227679e-02, -5.32227679e-02]],
 
        [[-3.47047062e-02,  3.47047062e-02],
         [-3.35509800e-02,  3.35509800e-02],
         [-4.53754719e-02,  4.53754719e-02],
         [-2.15031797e-02,  2.15031797e-02],
         [-2.45371369e-02,  2.45371369e-02],
         [-3.91520545e-02,  3.91520545e-02]],
 
        [[-5.37958443e-02,  5.37958443e-02],
         [-4.51532297e-02,  4.51532297e-02],
         [-3.72407464e-02,  3.72407464e-02],
         [ 3.05808276e-02, -3.05808276e-02],
         [-7.80999511e-02,  7.80999511e-02],
         [ 4.88541453e-03, -4.88541453e-03]],
 
        [[-4.47443764e-02,  4.47443764e-02],
         [ 2.62466008e-02, -2.62466008e-02],
         [-4.87473783e-02,  4.87473783e-02],
         [ 9.10437273e-04, -9.10437273e-04],
         [-7.02885080e-02,  7.02885080e-02],
         [-3.22003047e-02,  3.22003047e-02]],
 
        [[-3.07755956e-03,  3.07755956e-03],
         [-3.54187408e-02,  3.54187408e-02],
         [-4.72570700e-02,  4.72570700e-02],
         [-1.68429860e-02,  1.68429860e-02],
         [-7.29453002e-02,  7.29453002e-02],
         [-3.32818729e-02,  3.32818729e-02]],
 
        [[-4.64749023e-02,  4.64749023e-02],
         [-3.18401334e-02,  3.18401334e-02],
         [-3.92009498e-02,  3.92009498e-02],
         [-1.09952501e-02,  1.09952501e-02],
         [-5.01907736e-02,  5.01907736e-02],
         [-3.01215201e-02,  3.01215201e-02]],
 
        [[ 1.35865025e-02, -1.35865025e-02],
         [-3.36609858e-02,  3.36609858e-02],
         [-4.74008343e-02,  4.74008343e-02],
         [-2.21849526e-02,  2.21849526e-02],
         [-3.84845992e-02,  3.84845992e-02],
         [-5.06786600e-02,  5.06786600e-02]],
 
        [[-5.36768690e-02,  5.36768690e-02],
         [ 5.36719526e-03, -5.36719526e-03],
         [-3.83688579e-02,  3.83688579e-02],
         [-4.37552126e-03,  4.37552126e-03],
         [-5.43662371e-02,  5.43662371e-02],
         [-3.34032395e-02,  3.34032395e-02]],
 
        [[ 1.38013819e-01, -1.38013819e-01],
         [ 8.42321051e-02, -8.42321051e-02],
         [ 1.74314743e-01, -1.74314743e-01],
         [ 4.19096688e-02, -4.19096688e-02],
         [ 2.37474833e-01, -2.37474833e-01],
         [ 1.05231302e-01, -1.05231302e-01]],
 
        [[ 9.55114714e-02, -9.55114714e-02],
         [ 2.21973319e-01, -2.21973319e-01],
         [-8.10532836e-04,  8.10532836e-04],
         [ 1.28095656e-02, -1.28095656e-02],
         [ 3.25826099e-01, -3.25826099e-01],
         [ 1.58665484e-02, -1.58665484e-02]],
 
        [[-2.97952034e-03,  2.97952034e-03],
         [-3.49266773e-02,  3.49266773e-02],
         [-4.72570700e-02,  4.72570700e-02],
         [-1.52743585e-02,  1.52743585e-02],
         [-6.92557577e-02,  6.92557577e-02],
         [-3.91301456e-02,  3.91301456e-02]],
 
        [[-8.76757444e-03,  8.76757444e-03],
         [-3.96405330e-02,  3.96405330e-02],
         [-4.50008400e-02,  4.50008400e-02],
         [-9.40491662e-03,  9.40491662e-03],
         [-9.68369103e-03,  9.68369103e-03],
         [-4.63259744e-02,  4.63259744e-02]],
 
        [[-4.90025524e-02,  4.90025524e-02],
         [-3.22051823e-02,  3.22051823e-02],
         [-3.89613470e-02,  3.89613470e-02],
         [-1.10132810e-02,  1.10132810e-02],
         [-4.62538721e-02,  4.62538721e-02],
         [-3.13872945e-02,  3.13872945e-02]],
 
        [[ 1.29055622e-01, -1.29055622e-01],
         [-5.02155873e-02,  5.02155873e-02],
         [-4.21503139e-02,  4.21503139e-02],
         [-3.70542884e-02,  3.70542884e-02],
         [-4.81081558e-02,  4.81081558e-02],
         [-5.03508055e-02,  5.03508055e-02]]]),
 'X_sample':     percent_college  semi_urban_pct  median_household_income  percent_poverty  \
 43        28.832720        0.484252                    64044             13.6   
 40        30.288157        0.413043                    56360             13.9   
 46        26.959316        0.240602                    76471              9.9   
 12        35.848404        0.500000                    60830             11.0   
 24        32.042648        0.536585                    45928             19.5   
 31        31.799431        0.606061                    52021             17.5   
 17        29.152100        0.408333                    52256             16.0   
 32        24.268467        0.370968                    72038             13.1   
 3         29.507084        0.560000                    49020             16.0   
 30        22.916826        0.000000                    85786              9.1   
 13        28.604910        0.509804                    69212             11.4   
 8         15.547361        0.000000                    90395             14.1   
 49        31.452776        0.458333                    64177             10.4   
 6         24.491169        0.125000                    78920              9.9   
 47        33.307690        0.333333                    78674              9.8   
 4         28.893970        0.293103                    80423             11.8   
 36        31.172707        0.558442                    54447             15.1   
 33        30.872301        0.380000                    57388             13.6   
 19        29.324507        0.562500                    58824             10.9   
 48        25.967607        0.418182                    48659             16.2   
 15        32.542377        0.585859                    61807             11.0   
 9         29.736067        0.313433                    59198             12.7   
 16        31.660761        0.419048                    62028             11.3   
 26        32.688599        0.392857                    57248             12.6   
 44        35.426731        0.482759                    75705              8.8   
 25        30.086369        0.443478                    57375             12.9   
 11        31.638309        0.400000                    83734              9.0   
 0         29.912098        0.402985                    51771             15.6   
 45        25.852291        0.571429                    63293             10.1   
 27        33.386356        0.311828                    63290              9.9   
 34        36.171799        0.188679                    67402             10.5   
 5         29.465919        0.421875                    77104              9.4   
 29        28.643383        0.700000                    78571              7.5   
 37        34.312252        0.500000                    66955             11.5   
 1         35.292122        0.310345                    77203             10.2   
 21        23.049395        0.214286                    85700              9.5   
 2         33.813610        0.466667                    62027             13.5   
 39        26.346718        0.000000                    70383             11.6   
 35        29.063951        0.545455                    58704             13.0   
 23        32.411053        0.471264                    74529              8.9   
 41        32.668640        0.242424                    60414             11.9   
 10        28.107138        0.396226                    61950             13.5   
 22        32.799744        0.518072                    59522             12.9   
 18        27.149839        0.375000                    51108             18.8   
 50        36.730377        0.739130                    66152              9.9   
 20        25.420776        0.208333                    86644              9.1   
 7         26.731159        0.000000                    70348             11.2   
 42        28.035902        0.389474                    56047             13.8   
 14        28.955311        0.467391                    57617             11.9   
 28        33.873901        0.529412                    63268             12.7   
 38        24.395906        0.388060                    63455             12.0   
 
     rural_pct  percent_bachelor  
 43   0.192913         29.896368  
 40   0.021739         28.119732  
 46   0.157895         38.781448  
 12   0.227273         27.568047  
 24   0.256098         22.025591  
 31   0.181818         27.328411  
 17   0.300000         24.215286  
 32   0.016129         36.574459  
 3    0.173333         23.027790  
 30   0.000000         39.713642  
 13   0.098039         34.652561  
 8    0.000000         58.540707  
 49   0.180556         30.116730  
 6    0.000000         39.284241  
 47   0.128205         36.020138  
 4    0.068966         33.925964  
 36   0.207792         25.520313  
 33   0.160000         31.255686  
 19   0.125000         31.812195  
 48   0.200000         20.614605  
 15   0.202020         28.566288  
 9    0.029851         29.879576  
 16   0.400000         33.384693  
 26   0.517857         32.029671  
 44   0.172414         34.017944  
 25   0.260870         29.218016  
 11   0.000000         32.977276  
 0    0.164179         25.468332  
 45   0.214286         38.024567  
 27   0.548387         31.911806  
 34   0.698113         30.047791  
 5    0.312500         40.912342  
 29   0.000000         37.042934  
 37   0.138889         33.664604  
 1    0.586207         29.551214  
 21   0.000000         43.688908  
 2    0.000000         29.466806  
 39   0.000000         34.199402  
 35   0.022727         28.277195  
 23   0.218391         36.081844  
 41   0.636364         28.840324  
 10   0.138365         31.323006  
 22   0.168675         29.135920  
 18   0.078125         24.113539  
 50   0.173913         27.362514  
 20   0.000000         40.172043  
 7    0.000000         31.993366  
 42   0.168421         27.338949  
 14   0.054348         26.456558  
 28   0.235294         24.731855  
 38   0.059701         31.429663  }

!jupyter nbconvert notebook_states.ipynb --to html

[NbConvertApp] Converting notebook notebook_states.ipynb to html
[NbConvertApp] WARNING | Alternative text is missing on 28 image(s).
[NbConvertApp] Writing 49395640 bytes to notebook_states.html

	id	per_gop	per_dem	total_2016	dem_2016	gop_2016	percent_no_highschool	percent_highschool	percent_college	percent_bachelor	...	uic_4	uic_5	uic_6	uic_7	uic_8	uic_9	uic_10	uic_11	uic_12	target
count	51.000000	51.000000	51.000000	5.100000e+01	5.100000e+01	5.100000e+01	51.000000	51.000000	51.000000	51.000000	...	51.000000	51.000000	51.000000	51.000000	51.000000	51.000000	51.000000	51.000000	51.000000	51.000000
mean	28960.784314	0.586317	0.392727	2.495477e+06	1.193607e+06	1.180349e+06	10.461532	28.010589	29.757235	31.770644	...	0.034489	0.073865	0.088338	0.042785	0.092530	0.049669	0.050380	0.045169	0.052457	0.784314
std	15832.827649	0.146841	0.145555	2.386844e+06	1.278957e+06	1.076152e+06	2.706207	4.173287	4.023961	6.427930	...	0.037211	0.054218	0.079990	0.043547	0.087541	0.055531	0.066516	0.063557	0.084690	0.415390
min	1000.000000	0.053973	0.198169	0.000000e+00	0.000000e+00	0.000000e+00	6.449651	16.835115	15.547361	20.614605	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	16500.000000	0.503118	0.279586	7.360890e+05	2.703425e+05	3.713010e+05	8.253462	25.779302	27.592871	27.843889	...	0.000000	0.038075	0.032998	0.000000	0.029963	0.000000	0.000000	0.000000	0.000000	1.000000
50%	29000.000000	0.602136	0.376106	1.922218e+06	7.795350e+05	9.479340e+05	9.796140	28.003729	29.736067	31.255686	...	0.024096	0.071429	0.071429	0.034483	0.068966	0.031579	0.034483	0.027778	0.017241	1.000000
75%	41500.000000	0.693200	0.462139	3.088076e+06	1.680510e+06	1.545866e+06	12.696142	30.719540	32.605509	34.425982	...	0.057821	0.106456	0.128571	0.058586	0.136111	0.078030	0.068966	0.069048	0.065979	1.000000
max	56000.000000	0.780742	0.921497	9.631972e+06	5.931283e+06	4.681590e+06	16.692171	40.320034	36.730377	58.540707	...	0.150376	0.214286	0.375000	0.181818	0.400000	0.260870	0.303030	0.344828	0.379310	1.000000

PROJET DE SOUTENANCE - DATAGONG x 10.000 Codeurs¶

0. PRESENTATION¶

1. Setup¶

1.1. INSTALLATION DES DEPENDANCES¶

1.2. IMPORTATION DES LIBRAIRIES¶

2. IMPORTATION DES DONNÉES¶

3. PREPARATION ET CONSTITUTION DES DONNEES¶

3.1. Etape 1¶

3.2. Etape 2¶

3.3. Etape 3¶

3.4. Etape 4¶

REMARQUE¶

4. ANALYSE EXPLORATOIRE¶

4.1. Recuperation , Inspection et Creation de la variable cible (Target)¶

4.2. Distribution des votes pour les différents partis (Democrates et Republicains)¶

4.3. Analyses Univariées¶

4.4. Analyses Bivariées¶

5. MODELISATION¶

5.1. CheckUps¶

Cela peut poser problème, car un modèle de classification risque de favoriser la classe majoritaire et mal prédire les Démocrates.¶

5.2. Séparation features et target¶

5.3. Encodage des variables catégorielles¶

5.4. Modélisation proprement dite¶

6. EVALUATION¶

Analyse de la capacité de généralisation des modèles¶

Résultats observés¶

Accuracy¶

Recall¶

F1-Score¶

Interprétation¶

Conclusion¶

Procédons l'analyse des parametres pour voir les plus importantes¶

7. Exports vers le format HTML¶

	fips	county_name	state_name	state_code
0	1001	Autauga County	Alabama	NaN
1	1003	Baldwin County	Alabama	NaN
2	1005	Barbour County	Alabama	NaN
3	1007	Bibb County	Alabama	NaN
4	1009	Blount County	Alabama	NaN
...	...	...	...	...
19283	72145	Vega Baja Municipio, PR	NaN	PR
19284	72147	Vieques Municipio, PR	NaN	PR
19285	72149	Villalba Municipio, PR	NaN	PR
19286	72151	Yabucoa Municipio, PR	NaN	PR
19287	72153	Yauco Municipio, PR	NaN	PR

	fips	county_name	state_name	state_code
0	01001	Autauga County	Alabama	NaN
1	01003	Baldwin County	Alabama	NaN
2	01005	Barbour County	Alabama	NaN
3	01007	Bibb County	Alabama	NaN
4	01009	Blount County	Alabama	NaN
...	...	...	...	...
19283	72145	Vega Baja Municipio, PR	NaN	PR
19284	72147	Vieques Municipio, PR	NaN	PR
19285	72149	Villalba Municipio, PR	NaN	PR
19286	72151	Yabucoa Municipio, PR	NaN	PR
19287	72153	Yauco Municipio, PR	NaN	PR

	fips	county_name	state_name	state_code
0	00000	United States	NaN	US
1	01000	Alabama	NaN	AL
2	01001	Autauga County	Alabama	NaN
3	01003	Baldwin County	Alabama	NaN
4	01005	Barbour County	Alabama	NaN
...	...	...	...	...
3319	72145	Vega Baja Municipio, Puerto Rico	NaN	PR
3320	72147	Vieques Municipio, Puerto Rico	NaN	PR
3321	72149	Villalba Municipio, Puerto Rico	NaN	PR
3322	72151	Yabucoa Municipio, Puerto Rico	NaN	PR
3323	72153	Yauco Municipio, Puerto Rico	NaN	PR

	id	state_name	state_code	per_gop	per_dem	total_2016	dem_2016	gop_2016	percent_no_highschool	percent_highschool	...	uic_3	uic_4	uic_5	uic_6	uic_7	uic_8	uic_9	uic_10	uic_11	uic_12
3271	01000	Alabama	AL	0.647359	0.342648	2078165.0	718084.0	1306925.0	13.819302	30.800268	...	0.044776	0.059701	0.104478	0.208955	0.029851	0.000000	0.000000	0.044776	0.074627	0.000000
3272	02000	Alaska	AK	0.497797	0.420912	0.0	0.0	0.0	7.152934	28.003729	...	0.000000	0.000000	0.000000	0.000000	0.034483	0.068966	0.000000	0.068966	0.344828	0.379310
3273	04000	Arizona	AZ	0.548723	0.435861	2062810.0	936250.0	1021154.0	12.860705	23.858877	...	0.066667	0.066667	0.133333	0.066667	0.000000	0.066667	0.066667	0.000000	0.000000	0.000000
3274	05000	Arkansas	AR	0.688531	0.282032	1108615.0	378729.0	677904.0	13.430243	34.034885	...	0.040000	0.013333	0.053333	0.213333	0.040000	0.133333	0.146667	0.066667	0.026667	0.000000
3275	06000	California	CA	0.439389	0.537068	9631972.0	5931283.0	3184721.0	16.692171	20.487896	...	0.017241	0.051724	0.068966	0.086207	0.034483	0.051724	0.000000	0.000000	0.034483	0.017241

	id	state_name	state_code	per_gop	per_dem	total_2016	dem_2016	gop_2016	percent_no_highschool	percent_highschool	...	uic_4	uic_5	uic_6	uic_7	uic_8	uic_9	uic_10	uic_11	uic_12	target
0	1000	Alabama	AL	0.647359	0.342648	2078165	718084	1306925	13.819302	30.800268	...	0.059701	0.104478	0.208955	0.029851	0.000000	0.000000	0.044776	0.074627	0.000000	1
1	2000	Alaska	AK	0.497797	0.420912	0	0	0	7.152934	28.003729	...	0.000000	0.000000	0.000000	0.034483	0.068966	0.000000	0.068966	0.344828	0.379310	1
2	4000	Arizona	AZ	0.548723	0.435861	2062810	936250	1021154	12.860705	23.858877	...	0.066667	0.133333	0.066667	0.000000	0.066667	0.066667	0.000000	0.000000	0.000000	1
3	5000	Arkansas	AR	0.688531	0.282032	1108615	378729	677904	13.430243	34.034885	...	0.013333	0.053333	0.213333	0.040000	0.133333	0.146667	0.066667	0.026667	0.000000	1
4	6000	California	CA	0.439389	0.537068	9631972	5931283	3184721	16.692171	20.487896	...	0.051724	0.068966	0.086207	0.034483	0.051724	0.000000	0.000000	0.034483	0.017241	0

	percent_college	semi_urban_pct	median_household_income	percent_poverty	rural_pct	percent_bachelor	id_2000	id_4000	id_5000	id_6000	...
0	29.912098	0.402985	51771	15.6	0.164179	25.468332	0.0	0.0	0.0	0.0	...
1	35.292122	0.310345	77203	10.2	0.586207	29.551214	1.0	0.0	0.0	0.0	...
2	33.813610	0.466667	62027	13.5	0.000000	29.466806	0.0	1.0	0.0	0.0	...
3	29.507084	0.560000	49020	16.0	0.173333	23.027790	0.0	0.0	1.0	0.0	...
4	28.893970	0.293103	80423	11.8	0.068966	33.925964	0.0	0.0	0.0	1.0	...

	Model	Best Params	Train F1-Score	Test F1-Score	Train Accuracy	Test Accuracy	Train Recall	Test Recall	Train Precision	Test Precision
0	Logistic Regression	{'logreg__C': 0.1, 'logreg__solver': 'liblinear'}	1.0	0.909091	1.0	0.8750	1.0	0.833333	1.0	1.000000
1	Random Forest	{'rf__max_depth': 10, 'rf__min_samples_leaf': ...	1.0	0.960000	1.0	0.9375	1.0	1.000000	1.0	0.923077
2	XGBoost	{'xgb__colsample_bytree': 0.8, 'xgb__learning_...	1.0	0.869565	1.0	0.8125	1.0	0.833333	1.0	0.909091